feat: Marketing site

Converts Vantage from a single-admin self-hosted app into a multi-tenant
SaaS. Org isolation is org_id row-scoping in one shared deployment and
database; each org gets a subdomain via wildcard DNS, but the hostname is
a routing hint and never an authorization boundary.

- Orgs, users and roles (owner|admin|member) with local password auth,
  first-run bootstrap, and per-org OIDC replacing the global provider
- Redis sessions carrying org and role; host/session mismatch guard
- Every tenant-scoped collection carries org_id; handlers take org from
  the session only, never from client input
- Agent gRPC path resolves org from the servers record, so agent configs
  and the shared gRPC host are unchanged
- Per-org settings and ESO read token, replacing global singletons
- Frontend: login, setup and org settings pages; org-aware AuthProvider
- Migrations 0001-0003 backfill an existing single-tenant instance

Reviewed per task plus a whole-branch and a migration-focused pass.

.gitea/workflows/server-deploy.yml

@@ -35,3 +35,19 @@ jobs:
                     -t "$IMAGE" \
                     -f web/Dockerfile web/
                   docker push "$IMAGE"
+
+            - name: Build and push site image
+              run: |
+                  IMAGE="${{ vars.DOCKER_HOST }}/${{ github.repository_owner }}/vantage/site:latest"
+                  docker build \
+                    --build-arg NEXT_PUBLIC_SITE_API="${{ vars.SITE_API_URL }}" \
+                    --build-arg NEXT_PUBLIC_CONTACT_EMAIL="support@hostxtra.co.uk" \
+                    -t "$IMAGE" \
+                    -f site/Dockerfile site/
+                  docker push "$IMAGE"
+
+            - name: Build and push sitesvc image
+              run: |
+                  IMAGE="${{ vars.DOCKER_HOST }}/${{ github.repository_owner }}/vantage/sitesvc:latest"
+                  docker build -t "$IMAGE" -f sitesvc/Dockerfile sitesvc/
+                  docker push "$IMAGE"

.gitignore

@@ -8,4 +8,5 @@ installer/vantage-agent-windows-amd64.exe
 installer/*.msi
 installer/nssm.zip
 installer/checksums-msi.txt
-.next
+.next
+*.tsbuildinfo

agent/internal/checker/checker.go

@@ -0,0 +1,226 @@
+// Package checker runs service checks (http/tcp/icmp/tls) and returns a uniform
+// Result. It has no dependency on models or pb so it can be duplicated verbatim
+// into the agent module (agent-run monitors) — callers map their own monitor
+// representation onto Spec.
+package checker
+
+import (
+	"context"
+	"crypto/tls"
+	"fmt"
+	"io"
+	"net"
+	"net/http"
+	"os"
+	"strings"
+	"time"
+)
+
+// Check types (mirror models.Monitor* constants).
+const (
+	TypeHTTP = "http"
+	TypeTCP  = "tcp"
+	TypeICMP = "icmp"
+	TypeTLS  = "tls"
+)
+
+// Spec is a self-contained description of a single check.
+type Spec struct {
+	Type           string
+	URL            string
+	Host           string
+	Port           int
+	Method         string
+	ExpectedStatus int
+	Keyword        string
+	TLSWarnDays    int
+	Insecure       bool // skip TLS certificate verification (HTTP checks)
+	TimeoutSec     int
+}
+
+// Result is the uniform outcome of running a check.
+type Result struct {
+	Up         bool
+	LatencyMs  int
+	Message    string
+	CertExpiry *time.Time
+}
+
+func (s Spec) timeout() time.Duration {
+	t := s.TimeoutSec
+	if t <= 0 || t > 10 {
+		t = 10
+	}
+	return time.Duration(t) * time.Second
+}
+
+// Run executes the check described by s.
+func Run(ctx context.Context, s Spec) Result {
+	switch s.Type {
+	case TypeHTTP:
+		return runHTTP(ctx, s)
+	case TypeTCP:
+		return runTCP(ctx, s)
+	case TypeICMP:
+		return runICMP(ctx, s)
+	case TypeTLS:
+		return runTLS(ctx, s)
+	default:
+		return Result{Message: "unknown check type: " + s.Type}
+	}
+}
+
+func runHTTP(ctx context.Context, s Spec) Result {
+	method := s.Method
+	if method == "" {
+		method = http.MethodGet
+	}
+	expect := s.ExpectedStatus
+	if expect == 0 {
+		expect = 200
+	}
+	client := &http.Client{Timeout: s.timeout()}
+	if s.Insecure {
+		client.Transport = &http.Transport{TLSClientConfig: &tls.Config{InsecureSkipVerify: true}} //nolint:gosec // opt-in per monitor
+	}
+	start := time.Now()
+	req, err := http.NewRequestWithContext(ctx, method, s.URL, nil)
+	if err != nil {
+		return Result{Message: err.Error()}
+	}
+	resp, err := client.Do(req)
+	if err != nil {
+		return Result{LatencyMs: msSince(start), Message: err.Error()}
+	}
+	defer resp.Body.Close()
+	res := Result{LatencyMs: msSince(start), Up: true}
+	if resp.TLS != nil && len(resp.TLS.PeerCertificates) > 0 {
+		exp := resp.TLS.PeerCertificates[0].NotAfter
+		res.CertExpiry = &exp
+	}
+	if resp.StatusCode != expect {
+		return Result{LatencyMs: res.LatencyMs, CertExpiry: res.CertExpiry, Message: fmt.Sprintf("status %d (want %d)", resp.StatusCode, expect)}
+	}
+	if s.Keyword != "" {
+		body, _ := io.ReadAll(io.LimitReader(resp.Body, 1<<20))
+		if !strings.Contains(string(body), s.Keyword) {
+			return Result{LatencyMs: res.LatencyMs, CertExpiry: res.CertExpiry, Message: "keyword not found"}
+		}
+	}
+	return res
+}
+
+func runTCP(ctx context.Context, s Spec) Result {
+	addr := net.JoinHostPort(s.Host, fmt.Sprint(s.Port))
+	start := time.Now()
+	d := net.Dialer{Timeout: s.timeout()}
+	conn, err := d.DialContext(ctx, "tcp", addr)
+	if err != nil {
+		return Result{LatencyMs: msSince(start), Message: err.Error()}
+	}
+	conn.Close()
+	return Result{Up: true, LatencyMs: msSince(start)}
+}
+
+func runTLS(ctx context.Context, s Spec) Result {
+	port := s.Port
+	if port == 0 {
+		port = 443
+	}
+	addr := net.JoinHostPort(s.Host, fmt.Sprint(port))
+	start := time.Now()
+	d := net.Dialer{Timeout: s.timeout()}
+	conn, err := tls.DialWithDialer(&d, "tcp", addr, &tls.Config{ServerName: s.Host})
+	if err != nil {
+		return Result{LatencyMs: msSince(start), Message: err.Error()}
+	}
+	defer conn.Close()
+	certs := conn.ConnectionState().PeerCertificates
+	if len(certs) == 0 {
+		return Result{LatencyMs: msSince(start), Message: "no peer certificate"}
+	}
+	exp := certs[0].NotAfter
+	res := Result{LatencyMs: msSince(start), CertExpiry: &exp}
+	warn := s.TLSWarnDays
+	if warn <= 0 {
+		warn = 14
+	}
+	remaining := time.Until(exp)
+	if remaining <= 0 {
+		res.Message = "certificate expired"
+		return res
+	}
+	if remaining <= time.Duration(warn)*24*time.Hour {
+		res.Message = fmt.Sprintf("certificate expires in %d days", int(remaining.Hours()/24))
+		return res
+	}
+	res.Up = true
+	return res
+}
+
+func msSince(t time.Time) int { return int(time.Since(t).Milliseconds()) }
+
+// runICMP sends a single ICMP echo request and waits for the reply. Requires
+// raw-socket privileges (the agent and server run as root). Returns down with a
+// descriptive message when the socket cannot be opened or no reply arrives.
+func runICMP(ctx context.Context, s Spec) Result {
+	dst, err := net.ResolveIPAddr("ip4", s.Host)
+	if err != nil {
+		return Result{Message: err.Error()}
+	}
+	conn, err := net.ListenPacket("ip4:icmp", "0.0.0.0")
+	if err != nil {
+		return Result{Message: "icmp socket: " + err.Error()}
+	}
+	defer conn.Close()
+
+	id := os.Getpid() & 0xffff
+	pkt := icmpEcho(id, 1)
+	deadline := time.Now().Add(s.timeout())
+	if d, ok := ctx.Deadline(); ok && d.Before(deadline) {
+		deadline = d
+	}
+	_ = conn.SetDeadline(deadline)
+
+	start := time.Now()
+	if _, err := conn.WriteTo(pkt, dst); err != nil {
+		return Result{Message: err.Error()}
+	}
+	reply := make([]byte, 1500)
+	for {
+		n, peer, err := conn.ReadFrom(reply)
+		if err != nil {
+			return Result{LatencyMs: msSince(start), Message: "no reply"}
+		}
+		// Skip the IPv4 header (20 bytes) to reach the ICMP message.
+		if n < 28 || peer.String() != dst.String() {
+			continue
+		}
+		if reply[20] == 0 { // ICMP echo reply type
+			return Result{Up: true, LatencyMs: msSince(start)}
+		}
+	}
+}
+
+func icmpEcho(id, seq int) []byte {
+	// Type(8)=echo request, Code=0, Checksum, ID, Seq, no payload.
+	b := []byte{8, 0, 0, 0, byte(id >> 8), byte(id), byte(seq >> 8), byte(seq)}
+	cs := icmpChecksum(b)
+	b[2] = byte(cs >> 8)
+	b[3] = byte(cs)
+	return b
+}
+
+func icmpChecksum(b []byte) uint16 {
+	var sum uint32
+	for i := 0; i < len(b)-1; i += 2 {
+		sum += uint32(b[i])<<8 | uint32(b[i+1])
+	}
+	if len(b)%2 == 1 {
+		sum += uint32(b[len(b)-1]) << 8
+	}
+	for sum>>16 != 0 {
+		sum = (sum & 0xffff) + (sum >> 16)
+	}
+	return ^uint16(sum)
+}

agent/internal/grpc/client.go

@@ -126,6 +126,30 @@ func (c *Client) ReportUpdates(serverID, agentToken string, updates []pb.Package
 	return err
 }
 
+func (c *Client) ReportInventory(report *pb.InventoryReport) error {
+	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+	defer cancel()
+	_, err := c.client.ReportInventory(ctx, report)
+	return err
+}
+
+func (c *Client) SyncMonitors(serverID, agentToken string) ([]pb.MonitorSpec, error) {
+	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+	defer cancel()
+	resp, err := c.client.SyncMonitors(ctx, &pb.SyncMonitorsRequest{ServerId: serverID, AgentToken: agentToken})
+	if err != nil {
+		return nil, err
+	}
+	return resp.Monitors, nil
+}
+
+func (c *Client) ReportChecks(serverID, agentToken string, results []pb.CheckResult) error {
+	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+	defer cancel()
+	_, err := c.client.ReportChecks(ctx, &pb.ReportChecksRequest{ServerId: serverID, AgentToken: agentToken, Results: results})
+	return err
+}
+
 // CommandStream opens a long-lived bidirectional stream for server-pushed commands.
 // The caller controls the stream lifetime via ctx.
 func (c *Client) CommandStream(ctx context.Context) (pb.Vantage_CommandStreamClient, error) {

agent/internal/grpc/pb/vantage.pb.go

@@ -60,6 +60,75 @@ type ReportUpdatesRequest struct {
 
 type ReportUpdatesResponse struct{}
 
+// Inventory report message types
+
+type CPUReport struct {
+	Model    string  `json:"model,omitempty"`
+	Cores    int     `json:"cores,omitempty"`
+	UsagePct float64 `json:"usage_pct"`
+	Load1    float64 `json:"load1,omitempty"`
+}
+type MemReport struct {
+	TotalBytes uint64 `json:"total_bytes"`
+	UsedBytes  uint64 `json:"used_bytes"`
+}
+type PartitionReport struct {
+	Device     string `json:"device"`
+	Mountpoint string `json:"mountpoint"`
+	Fstype     string `json:"fstype,omitempty"`
+	TotalBytes uint64 `json:"total_bytes"`
+	UsedBytes  uint64 `json:"used_bytes"`
+}
+type InventoryReport struct {
+	ServerId      string            `json:"server_id"`
+	AgentToken    string            `json:"agent_token"`
+	IncludeStatic bool              `json:"include_static"`
+	CPU           *CPUReport        `json:"cpu,omitempty"`
+	Memory        *MemReport        `json:"memory,omitempty"`
+	SwapTotal     uint64            `json:"swap_total"`
+	SwapUsed      uint64            `json:"swap_used"`
+	Partitions    []PartitionReport `json:"partitions,omitempty"`
+	Kernel        string            `json:"kernel,omitempty"`
+}
+type InventoryReportResponse struct{}
+
+// Monitor sync / check report message types
+
+type MonitorSpec struct {
+	MonitorId      string `json:"monitor_id"`
+	Type           string `json:"type"`
+	URL            string `json:"url,omitempty"`
+	Host           string `json:"host,omitempty"`
+	Port           int    `json:"port,omitempty"`
+	Method         string `json:"method,omitempty"`
+	ExpectedStatus int    `json:"expected_status,omitempty"`
+	Keyword        string `json:"keyword,omitempty"`
+	TLSWarnDays    int    `json:"tls_warn_days,omitempty"`
+	Insecure       bool   `json:"insecure,omitempty"`
+	IntervalSec    int    `json:"interval_sec"`
+	Retries        int    `json:"retries"`
+}
+type SyncMonitorsRequest struct {
+	ServerId   string `json:"server_id"`
+	AgentToken string `json:"agent_token"`
+}
+type SyncMonitorsResponse struct {
+	Monitors []MonitorSpec `json:"monitors,omitempty"`
+}
+type CheckResult struct {
+	MonitorId      string `json:"monitor_id"`
+	Up             bool   `json:"up"`
+	LatencyMs      int    `json:"latency_ms"`
+	Message        string `json:"message,omitempty"`
+	CertExpiryUnix int64  `json:"cert_expiry_unix,omitempty"`
+}
+type ReportChecksRequest struct {
+	ServerId   string        `json:"server_id"`
+	AgentToken string        `json:"agent_token"`
+	Results    []CheckResult `json:"results,omitempty"`
+}
+type ReportChecksResponse struct{}
+
 type ApplyUpdatesCmd struct{}
 
 type ServerCommand struct {
@@ -190,6 +259,9 @@ type VantageClient interface {
 	SyncKeys(ctx context.Context, in *SyncRequest, opts ...grpc.CallOption) (*SyncResponse, error)
 	UploadGeneratedKey(ctx context.Context, in *UploadKeyRequest, opts ...grpc.CallOption) (*UploadKeyResponse, error)
 	ReportUpdates(ctx context.Context, in *ReportUpdatesRequest, opts ...grpc.CallOption) (*ReportUpdatesResponse, error)
+	ReportInventory(ctx context.Context, in *InventoryReport, opts ...grpc.CallOption) (*InventoryReportResponse, error)
+	SyncMonitors(ctx context.Context, in *SyncMonitorsRequest, opts ...grpc.CallOption) (*SyncMonitorsResponse, error)
+	ReportChecks(ctx context.Context, in *ReportChecksRequest, opts ...grpc.CallOption) (*ReportChecksResponse, error)
 	CommandStream(ctx context.Context, opts ...grpc.CallOption) (Vantage_CommandStreamClient, error)
 }
 
@@ -245,6 +317,30 @@ func (c *keyManagerClient) ReportUpdates(ctx context.Context, in *ReportUpdatesR
 	return out, nil
 }
 
+func (c *keyManagerClient) ReportInventory(ctx context.Context, in *InventoryReport, opts ...grpc.CallOption) (*InventoryReportResponse, error) {
+	out := new(InventoryReportResponse)
+	if err := c.cc.Invoke(ctx, "/vantage.v1.Vantage/ReportInventory", in, out, opts...); err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
+func (c *keyManagerClient) SyncMonitors(ctx context.Context, in *SyncMonitorsRequest, opts ...grpc.CallOption) (*SyncMonitorsResponse, error) {
+	out := new(SyncMonitorsResponse)
+	if err := c.cc.Invoke(ctx, "/vantage.v1.Vantage/SyncMonitors", in, out, opts...); err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
+func (c *keyManagerClient) ReportChecks(ctx context.Context, in *ReportChecksRequest, opts ...grpc.CallOption) (*ReportChecksResponse, error) {
+	out := new(ReportChecksResponse)
+	if err := c.cc.Invoke(ctx, "/vantage.v1.Vantage/ReportChecks", in, out, opts...); err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
 func (c *keyManagerClient) CommandStream(ctx context.Context, opts ...grpc.CallOption) (Vantage_CommandStreamClient, error) {
 	desc := &grpc.StreamDesc{StreamName: "CommandStream", ServerStreams: true, ClientStreams: true}
 	stream, err := c.cc.NewStream(ctx, desc, "/vantage.v1.Vantage/CommandStream", opts...)

agent/internal/inventory/collect_linux.go

@@ -0,0 +1,157 @@
+//go:build linux
+
+package inventory
+
+import (
+	"bufio"
+	"os"
+	"strconv"
+	"strings"
+	"syscall"
+	"time"
+
+	"github.com/mrhid6/vantage/agent/internal/grpc/pb"
+)
+
+func collect(r *pb.InventoryReport, includeStatic bool) {
+	r.CPU.UsagePct = cpuUsage()
+	r.CPU.Load1 = load1()
+	memTotal, memAvail, swapTotal, swapFree := meminfo()
+	if memTotal > memAvail {
+		r.Memory.UsedBytes = memTotal - memAvail
+	}
+	if swapTotal > swapFree {
+		r.SwapUsed = swapTotal - swapFree
+	}
+	if includeStatic {
+		r.Memory.TotalBytes = memTotal
+		r.SwapTotal = swapTotal
+		r.CPU.Model, r.CPU.Cores = cpuStatic()
+		r.Kernel = kernel()
+		r.Partitions = partitions()
+	}
+}
+
+func readProc(path string) string { b, _ := os.ReadFile(path); return string(b) }
+
+func cpuSample() (idle, total uint64) {
+	f, err := os.Open("/proc/stat")
+	if err != nil {
+		return
+	}
+	defer f.Close()
+	sc := bufio.NewScanner(f)
+	if sc.Scan() {
+		fields := strings.Fields(sc.Text()) // cpu user nice system idle iowait ...
+		for i, v := range fields[1:] {
+			n, _ := strconv.ParseUint(v, 10, 64)
+			total += n
+			if i == 3 { // idle
+				idle = n
+			}
+		}
+	}
+	return
+}
+
+func cpuUsage() float64 {
+	i1, t1 := cpuSample()
+	time.Sleep(100 * time.Millisecond)
+	i2, t2 := cpuSample()
+	dt := float64(t2 - t1)
+	if dt <= 0 {
+		return 0
+	}
+	return (1 - float64(i2-i1)/dt) * 100
+}
+
+func load1() float64 {
+	fields := strings.Fields(readProc("/proc/loadavg"))
+	if len(fields) > 0 {
+		v, _ := strconv.ParseFloat(fields[0], 64)
+		return v
+	}
+	return 0
+}
+
+func meminfo() (total, avail, swapTotal, swapFree uint64) {
+	f, err := os.Open("/proc/meminfo")
+	if err != nil {
+		return
+	}
+	defer f.Close()
+	sc := bufio.NewScanner(f)
+	for sc.Scan() {
+		fields := strings.Fields(sc.Text())
+		if len(fields) < 2 {
+			continue
+		}
+		kb, _ := strconv.ParseUint(fields[1], 10, 64)
+		b := kb * 1024
+		switch strings.TrimSuffix(fields[0], ":") {
+		case "MemTotal":
+			total = b
+		case "MemAvailable":
+			avail = b
+		case "SwapTotal":
+			swapTotal = b
+		case "SwapFree":
+			swapFree = b
+		}
+	}
+	return
+}
+
+func cpuStatic() (model string, cores int) {
+	f, err := os.Open("/proc/cpuinfo")
+	if err != nil {
+		return
+	}
+	defer f.Close()
+	sc := bufio.NewScanner(f)
+	for sc.Scan() {
+		line := sc.Text()
+		if strings.HasPrefix(line, "processor") {
+			cores++
+		} else if strings.HasPrefix(line, "model name") && model == "" {
+			if i := strings.Index(line, ":"); i >= 0 {
+				model = strings.TrimSpace(line[i+1:])
+			}
+		}
+	}
+	return
+}
+
+func kernel() string {
+	return strings.TrimSpace(readProc("/proc/sys/kernel/osrelease"))
+}
+
+func partitions() []pb.PartitionReport {
+	allowed := map[string]bool{"ext4": true, "xfs": true, "btrfs": true, "zfs": true, "vfat": true, "ntfs": true, "ext3": true}
+	f, err := os.Open("/proc/mounts")
+	if err != nil {
+		return nil
+	}
+	defer f.Close()
+	var out []pb.PartitionReport
+	seen := map[string]bool{}
+	sc := bufio.NewScanner(f)
+	for sc.Scan() {
+		fields := strings.Fields(sc.Text())
+		if len(fields) < 3 || !allowed[fields[2]] || seen[fields[1]] {
+			continue
+		}
+		seen[fields[1]] = true
+		var st syscall.Statfs_t
+		if syscall.Statfs(fields[1], &st) != nil {
+			continue
+		}
+		total := st.Blocks * uint64(st.Bsize)
+		free := st.Bavail * uint64(st.Bsize)
+		out = append(out, pb.PartitionReport{
+			Device: fields[0], Mountpoint: fields[1], Fstype: fields[2],
+			TotalBytes: total, UsedBytes: total - free,
+		})
+	}
+	return out
+}

agent/internal/inventory/collect_other.go

@@ -0,0 +1,8 @@
+//go:build !linux
+
+package inventory
+
+import "github.com/mrhid6/vantage/agent/internal/grpc/pb"
+
+// collect is a no-op best-effort stub on non-Linux platforms.
+func collect(r *pb.InventoryReport, includeStatic bool) {}

agent/internal/inventory/inventory.go

@@ -0,0 +1,11 @@
+package inventory
+
+import "github.com/mrhid6/vantage/agent/internal/grpc/pb"
+
+// Collect gathers metrics always and static hardware info when includeStatic.
+// Platform specifics are provided by collect_linux.go / collect_other.go.
+func Collect(includeStatic bool) *pb.InventoryReport {
+	r := &pb.InventoryReport{IncludeStatic: includeStatic, CPU: &pb.CPUReport{}, Memory: &pb.MemReport{}}
+	collect(r, includeStatic)
+	return r
+}

agent/internal/monitors/monitors.go

@@ -0,0 +1,166 @@
+// Package monitors runs agent-side service checks. It polls the server for the
+// monitors assigned to this agent (SyncMonitors), runs each on its own interval
+// using the local checker package, and reports results back (ReportChecks).
+package monitors
+
+import (
+	"context"
+	"log"
+	"sync"
+	"time"
+
+	"github.com/mrhid6/vantage/agent/internal/checker"
+	"github.com/mrhid6/vantage/agent/internal/config"
+	grpcclient "github.com/mrhid6/vantage/agent/internal/grpc"
+	"github.com/mrhid6/vantage/agent/internal/grpc/pb"
+)
+
+// syncInterval controls how often the agent re-fetches its assigned monitors.
+const syncInterval = 30 * time.Second
+
+type runner struct {
+	intervalSec int
+	cancel      context.CancelFunc
+}
+
+// Run starts the agent monitor loop and blocks until ctx is cancelled.
+func Run(ctx context.Context, cfg *config.Config) {
+	active := map[string]*runner{}
+	var mu sync.Mutex
+
+	// results is a shared channel every check writes to; a single reporter
+	// goroutine batches and ships them so we make one ReportChecks call per tick.
+	results := make(chan pb.CheckResult, 64)
+	go reporter(ctx, cfg, results)
+
+	syncOnce := func() {
+		specs, err := fetchSpecs(cfg)
+		if err != nil {
+			log.Printf("monitors: sync: %v", err)
+			return
+		}
+		want := map[string]pb.MonitorSpec{}
+		for _, s := range specs {
+			want[s.MonitorId] = s
+		}
+
+		mu.Lock()
+		defer mu.Unlock()
+		for id, r := range active {
+			s, ok := want[id]
+			if !ok || s.IntervalSec != r.intervalSec {
+				r.cancel()
+				delete(active, id)
+			}
+		}
+		for id, s := range want {
+			if _, ok := active[id]; ok {
+				continue
+			}
+			rctx, cancel := context.WithCancel(ctx)
+			active[id] = &runner{intervalSec: s.IntervalSec, cancel: cancel}
+			go runSpec(rctx, s, results)
+		}
+	}
+
+	syncOnce()
+	t := time.NewTicker(syncInterval)
+	defer t.Stop()
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-t.C:
+			syncOnce()
+		}
+	}
+}
+
+func fetchSpecs(cfg *config.Config) ([]pb.MonitorSpec, error) {
+	client, err := grpcclient.New(cfg.ServerURL, cfg.TLS)
+	if err != nil {
+		return nil, err
+	}
+	defer client.Close()
+	return client.SyncMonitors(cfg.ServerID, cfg.AgentToken)
+}
+
+func runSpec(ctx context.Context, s pb.MonitorSpec, out chan<- pb.CheckResult) {
+	interval := time.Duration(s.IntervalSec) * time.Second
+	if interval <= 0 {
+		interval = 60 * time.Second
+	}
+	spec := checker.Spec{
+		Type:           s.Type,
+		URL:            s.URL,
+		Host:           s.Host,
+		Port:           s.Port,
+		Method:         s.Method,
+		ExpectedStatus: s.ExpectedStatus,
+		Keyword:        s.Keyword,
+		TLSWarnDays:    s.TLSWarnDays,
+		Insecure:       s.Insecure,
+		TimeoutSec:     s.IntervalSec,
+	}
+
+	run := func() {
+		res := checker.Run(ctx, spec)
+		cr := pb.CheckResult{MonitorId: s.MonitorId, Up: res.Up, LatencyMs: res.LatencyMs, Message: res.Message}
+		if res.CertExpiry != nil {
+			cr.CertExpiryUnix = res.CertExpiry.Unix()
+		}
+		select {
+		case out <- cr:
+		case <-ctx.Done():
+		}
+	}
+
+	run()
+	t := time.NewTicker(interval)
+	defer t.Stop()
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-t.C:
+			run()
+		}
+	}
+}
+
+// reporter batches results on a short interval and ships each batch in one call.
+func reporter(ctx context.Context, cfg *config.Config, in <-chan pb.CheckResult) {
+	t := time.NewTicker(5 * time.Second)
+	defer t.Stop()
+	var batch []pb.CheckResult
+	flush := func() {
+		if len(batch) == 0 {
+			return
+		}
+		client, err := grpcclient.New(cfg.ServerURL, cfg.TLS)
+		if err != nil {
+			log.Printf("monitors: report dial: %v", err)
+			batch = nil
+			return
+		}
+		if err := client.ReportChecks(cfg.ServerID, cfg.AgentToken, batch); err != nil {
+			log.Printf("monitors: report: %v", err)
+		}
+		client.Close()
+		batch = nil
+	}
+	for {
+		select {
+		case <-ctx.Done():
+			flush()
+			return
+		case r := <-in:
+			batch = append(batch, r)
+			if len(batch) >= 32 {
+				flush()
+			}
+		case <-t.C:
+			flush()
+		}
+	}
+}

agent/internal/sync/sync.go

@@ -21,7 +21,9 @@ import (
 	agentexec "github.com/mrhid6/vantage/agent/internal/exec"
 	grpcclient "github.com/mrhid6/vantage/agent/internal/grpc"
 	"github.com/mrhid6/vantage/agent/internal/grpc/pb"
+	"github.com/mrhid6/vantage/agent/internal/inventory"
 	"github.com/mrhid6/vantage/agent/internal/keys"
+	"github.com/mrhid6/vantage/agent/internal/monitors"
 	"github.com/mrhid6/vantage/agent/internal/updates"
 )
 
@@ -68,6 +70,12 @@ func Run(ctx context.Context, cfg *config.Config, version string) error {
 	// Check for OS updates on startup and then hourly.
 	go runUpdateCheck(ctx, cfg)
 
+	// Report host inventory: metrics every 30s, full static snapshot every 15 min.
+	go runInventory(ctx, cfg)
+
+	// Run agent-side service monitors assigned to this server.
+	go monitors.Run(ctx, cfg)
+
 	ticker := time.NewTicker(cfg.PollInterval)
 	defer ticker.Stop()
 
@@ -272,6 +280,40 @@ func runUpdateCheck(ctx context.Context, cfg *config.Config) {
 	}
 }
 
+// runInventory reports host metrics every 30s and a full static snapshot every
+// 15 min (and once immediately on startup so static fields populate without delay).
+func runInventory(ctx context.Context, cfg *config.Config) {
+	client, err := grpcclient.New(cfg.ServerURL, cfg.TLS)
+	if err != nil {
+		log.Printf("inventory dial error: %v", err)
+		return
+	}
+	defer client.Close()
+
+	report := func(static bool) {
+		r := inventory.Collect(static)
+		r.ServerId = cfg.ServerID
+		r.AgentToken = cfg.AgentToken
+		if err := client.ReportInventory(r); err != nil {
+			log.Printf("report inventory: %v", err)
+		}
+	}
+
+	report(true) // full snapshot on startup
+	ticker := time.NewTicker(30 * time.Second)
+	defer ticker.Stop()
+	tick := 0
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-ticker.C:
+			tick++
+			report(tick%30 == 0) // every 30th tick = 15 min → include static
+		}
+	}
+}
+
 func handleApplyUpdates(cfg *config.Config, cmd *pb.ServerCommand) {
 	log.Printf("applying OS updates (cmd=%s)…", cmd.CommandId)
 	if err := updates.ApplyAll(); err != nil {

claude.md

@@ -1,34 +1,38 @@
 # Vantage
 
-A self-hosted SSH key management system. A central server (Go + Next.js + MongoDB) manages public key assignments across servers. A lightweight Go agent runs on each managed server, polls the central server via gRPC, and atomically rewrites `/root/.ssh/authorized_keys` to match the desired state.
+A self-hosted, multi-tenant infrastructure control plane. It started as SSH key management and has grown into fleet management: SSH key assignment, workflow/script execution, service monitoring, a secrets vault, a browser console (SSH/RDP/VNC), and OS update management.
+
+A central server (Go + Next.js + MongoDB + Redis) drives a lightweight Go agent installed on each managed server. Agents poll over gRPC and also hold a bidirectional command stream for push-style commands.
 
 ---
 
 ## Architecture Overview
 
 ```
-┌─────────────────────────────────┐
-│  Next.js Frontend               │
-│  - Upload/manage keys           │
-│  - Add servers (install script) │
-│  - Assign/revoke per server     │
-└────────────┬────────────────────┘
-             │ REST
-┌────────────▼────────────────────┐
-│  Go Backend                     │
-│  - REST API for frontend        │
-│  - gRPC server for agents       │
-│  - MongoDB                      │
-└────────────┬────────────────────┘
-             │ gRPC (TLS)
-┌────────────▼────────────────────┐
-│  Go Agent (per server)          │
-│  - Polls every 30s              │
-│  - Rewrites authorized_keys     │
-│  - Can generate SSH keypairs    │
-└─────────────────────────────────┘
+┌──────────────────────────────────────────────┐
+│  Next.js 16 Frontend (web, :3000)            │
+│  servers · keys · workflows · monitors       │
+│  secrets · audit · console · settings        │
+└───────────────┬──────────────────────────────┘
+                │ REST + cookie session
+┌───────────────▼──────────────────────────────┐
+│  Go Backend (server)                         │
+│  :8080 REST (gin)   :9090 gRPC (agents)      │
+│  MongoDB (state) · Redis (sessions)          │
+│  monitor scheduler · workflow runner         │
+│  guacd tunnel proxy for browser console      │
+└───────────────┬──────────────────────────────┘
+                │ gRPC (TLS) — outbound from agent only
+┌───────────────▼──────────────────────────────┐
+│  Go Agent (per server, Linux + Windows)      │
+│  polls SyncKeys · CommandStream              │
+│  rewrites authorized_keys (Linux only)       │
+│  runs workflow steps · monitors · inventory  │
+└──────────────────────────────────────────────┘
 ```
 
+Multi-tenancy: every domain document carries `org_id`, and every service query is scoped by it. Org is resolved from the session, and optionally cross-checked against the request host (`<slug>.vantage.<tld>`).
+
 ---
 
 ## Repository Structure
@@ -36,264 +40,359 @@ A self-hosted SSH key management system. A central server (Go + Next.js + MongoD
 ```
 vantage/
 ├── agent/
-│   ├── cmd/main.go
+│   ├── cmd/main.go                  # flags: -generate-key
 │   └── internal/
-│       ├── config/
-│       ├── grpc/
-│       ├── keys/
-│       └── sync/
+│       ├── checker/                 # monitor check execution
+│       ├── config/                  # config.yaml load/save
+│       ├── exec/                    # workflow step execution
+│       ├── grpc/                    # client + generated pb
+│       ├── inventory/               # CPU/mem/disk collection (linux/other)
+│       ├── keys/                    # authorized_keys read/diff/write
+│       ├── monitors/                # agent-run monitor loop
+│       ├── sync/                    # poll loop + command stream
+│       └── updates/                 # OS package update check/apply
 ├── server/
 │   ├── cmd/main.go
 │   └── internal/
-│       ├── api/          # REST handlers for Next.js
-│       ├── grpc/         # gRPC server implementation
-│       ├── models/       # MongoDB models
-│       └── services/
-│           ├── keys.go
-│           ├── servers.go
-│           └── sync.go   # builds desired state per server
-├── web/
-│   ├── app/
-│   └── components/
-├── proto/
-│   └── vantage/v1/vantage.proto
-├── deploy/
-│   ├── docker-compose.yml
-│   └── agent.service
-└── .gitea/
-    └── workflows/
-        ├── agent-release.yml
-        └── server-deploy.yml
+│       ├── api/                     # REST handlers
+│       ├── auth/                    # local, OIDC, session, middleware, orghost
+│       ├── checker/                 # server-run monitor checks
+│       ├── db/                      # mongo connect + Col()
+│       ├── grpc/                    # gRPC server + generated pb
+│       ├── models/                  # MongoDB documents
+│       ├── monitorsched/            # server-side monitor scheduler
+│       ├── notify/                  # smtp, http, templating, dispatch
+│       └── services/                # business logic + migrations
+├── web/                             # the application UI (authenticated)
+│   ├── app/(app)/                   # authed routes
+│   ├── app/login, app/setup         # unauthed routes
+│   ├── components/                  # ui/, workflows/, monitors/, Sidebar
+│   └── lib/                         # api client, guac console, query client
+├── site/                            # public marketing site
+│   ├── app/                         # one directory per route
+│   ├── components/                  # Nav, Footer, Logo, InstrumentPanel, forms
+│   ├── assets/                      # image sources, not served
+│   └── Dockerfile                   # same shape as web/: standalone, node, 3000
+├── sitesvc/                         # public forms: contact mail + signup
+│   ├── cmd/main.go
+│   └── internal/
+│       ├── api/                     # contact, signup, verify
+│       ├── mail/                    # SMTP
+│       ├── models/                  # mirrors server org/user + pending signup
+│       ├── provision/               # slug rules mirrored from the control plane
+│       └── store/                   # mongo: pending signups, org/user creation
+├── proto/vantage/v1/vantage.proto
+├── installer/                       # Windows: setup.ps1, nssm.exe, WiX .wxs
+├── deploy/                          # docker-compose.yml, agent.service
+└── .gitea/workflows/                # agent-release.yml, server-deploy.yml
 ```
 
 ---
 
+## Subsystems
+
+### SSH keys
+Upload a public key, assign it per server, revoke softly. The agent diffs desired vs on-disk state and rewrites `/root/.ssh/authorized_keys` atomically. Keys can also be generated *on* a server by the agent; the private half can optionally be uploaded and is stored AES-256-GCM encrypted.
+
+### Workflows
+A library of reusable **steps** (bash or PowerShell scripts with declared inputs, outputs, and secret refs) composed into **workflows** targeting a set of servers. Running one snapshots the resolved steps into a `WorkflowRun`, then dispatches `RunStepCmd` over the agent command stream. Step stdout/stderr streams back as `StepOutputChunk` and is written to a log file on disk; the UI streams it live. Steps support `on_failure: stop|continue|retry`, per-run env passed between steps via `output_env`, and a per-run workspace directory the agent cleans up at the end.
+
+Default steps are seeded per org at boot (`SeedDefaultSteps`). Logs are swept by retention (`workflow_log_retention_days`; nil = 30 days, 0 = forever).
+
+### Monitors
+HTTP, TCP, ICMP and TLS checks. Each monitor has a `runner`: `"server"` (executed by the server-side scheduler) or a `server_id` (pushed to that agent, which runs it locally and reports results). Consecutive failures beyond `retries` flip state to `down`, open an `Incident`, and notify. Hourly `Rollup` documents back the uptime graphs.
+
+### Notification channels
+Per-org outbound destinations: `webhook`, `smtp`, `discord`, `slack`, `telegram`. Monitors reference channels by ID. Channels are testable from the UI.
+
+### Secrets vault
+Key/value pairs grouped by name, encrypted at rest with AES-256-GCM. Consumed two ways: referenced by workflow steps via `secret_refs` (injected as env at execution), and read by Kubernetes External Secrets Operator via `GET /api/secrets/:group/values` using a bearer token whose SHA-256 hash is stored in settings.
+
+### Browser console
+`POST /api/console/connect` mints a one-time session token; `GET /api/console/tunnel` upgrades to a WebSocket and proxies to **guacd** (Apache Guacamole daemon) using `github.com/wwt/guac`. SSH connections authenticate with a stored private key; RDP/VNC credentials are encrypted, single-use, and consumed when the tunnel opens.
+
+### Inventory and OS updates
+Agents report CPU/memory/swap/partitions/kernel — metrics every 30s, full static snapshot every 15 min. They also check for pending OS package updates hourly and can apply them on command (`ApplyUpdatesCmd`).
+
+### Agent self-update
+`UpdateAgentCmd` carries a target version and Gitea base URL; the agent downloads and replaces itself.
+
+### Marketing site and sitesvc
+`site/` is a separate Next.js app built exactly like `web/` — `output: "standalone"`, run by Node in a `node:26-alpine` image, listening on `3000` and published as `3001`. Both of its forms post to `sitesvc`; the control plane is not involved and has no public signup endpoint.
+
+`sitesvc/` (port `8082`) owns both flows end to end:
+
+| Form | Endpoint | Effect |
+| --- | --- | --- |
+| Contact | `POST /api/contact` | Emails `support@hostxtra.co.uk`, `Reply-To` the sender. Nothing stored. |
+| Create organisation | `POST /api/signup` | Records a pending signup and emails a verification link. |
+| Verification link | `GET /api/verify?token=…` | Creates the org and its owner, then redirects to `APP_LOGIN_URL`. |
+
+All three are deliberately **excluded from the self-hosted deployment**: `deploy/docker-compose.yml` mentions none of them, and they live in `deploy/docker-compose.site.yml` instead.
+
+```bash
+# self-hosted install — no marketing site, no sitesvc
+docker compose up -d
+
+# vantage.sh — control plane plus the public site
+docker compose -f docker-compose.yml -f docker-compose.site.yml up -d
+```
+
+### Signup and verification
+
+**Nothing is written to `orgs` or `users` until the emailed link is opened.** A signup lands in sitesvc's own `site_pending_signups` collection holding the org name, the address, and the password already bcrypt-hashed at cost 12. The consequence is worth stating: an address nobody controls can never occupy an email, hold an organisation slug, or produce an account that can sign in. It also means the control plane's login path needs no concept of "unverified".
+
+- The token is 32 random bytes; only its **SHA-256 hash** is stored, so a leaked database yields no working links.
+- `Verify` deletes the pending record **atomically before provisioning** (`FindOneAndDelete`), so a double-clicked link cannot create two organisations — the second delete matches nothing.
+- Links expire after 24 hours, and a **TTL index** lets Mongo drop abandoned signups so password hashes do not linger.
+- Re-submitting the form for the same address replaces the previous pending record, so only the newest link works.
+- If the owner insert fails after the org is created, the org is rolled back rather than stranded holding a slug. The rollback refuses to touch an org that has users.
+- Rate limited to 3 signups per client IP per hour, plus a honeypot field.
+
+### The one piece of duplicated logic
+
+`sitesvc/internal/provision` and `sitesvc/internal/models` mirror the control plane's slug rules, reserved names, bcrypt cost and document shapes. They are duplicated rather than imported because sitesvc is a separate module that deliberately does not depend on the server.
+
+**Nothing enforces the match automatically.** If the control plane's `Slugify`, `reservedSlugs`, `CreateOrg` or `CreateUser` change, update `sitesvc/internal/provision` in the same commit — a divergence would provision tenants under rules the app does not agree with.
+
+sitesvc also (re)declares the unique indexes on `users.email` and `orgs.slug` at boot so it does not depend on the server having started first. Creating an existing index is a no-op.
+
+---
+
+## Auth and Orgs
+
+- **Bootstrap** — first run has no users. `GET /auth/bootstrap-status` drives `/setup`, `POST /auth/bootstrap` creates the first org plus its owner.
+- **Local auth** — email + password (bcrypt), `POST /auth/login`.
+- **OIDC** — configured *per org* (`org_oidc`), issuer + client ID + encrypted client secret. `/auth/oidc/start` → `/auth/oidc/callback`.
+- **Sessions** — opaque 32-byte hex ID in the `km_session` cookie, session body stored in Redis with a 24h TTL.
+- **Roles** — `owner`, `admin`, `member`. `/api/settings` and `/api/org/*` require owner or admin.
+- **Host/org guard** — `APP_ROOT_LABEL` (default `vantage`) defines the app root label. A request to `<slug>.vantage.<tld>` resolves that org from the slug and rejects sessions belonging to a different one. Org lookups are cached for 60s.
+
+Unique indexes on user email and org slug are a **security property**, not an optimisation: `GetUserByEmail` does an unscoped `FindOne`, so duplicates would let the OIDC cross-org guard compare against an arbitrary user. Same for duplicate settings docs and duplicate ESO token hashes.
+
+---
+
 ## gRPC API
 
 ```protobuf
-syntax = "proto3";
-package vantage.v1;
-
 service Vantage {
-  rpc Register(RegisterRequest)           returns (RegisterResponse);
-  rpc SyncKeys(SyncRequest)               returns (SyncResponse);
-  rpc UploadGeneratedKey(UploadKeyRequest) returns (UploadKeyResponse);
-}
-
-message RegisterRequest {
-  string server_id      = 1;
-  string pre_reg_token  = 2;
-  string hostname       = 3;
-  string ip_address     = 4;
-  string os_info        = 5;
-}
-message RegisterResponse {
-  string agent_token = 1;
-}
-
-message SyncRequest {
-  string server_id   = 1;
-  string agent_token = 2;
-}
-message SyncResponse {
-  repeated string public_keys = 1; // full authorized_keys lines
-}
-
-message UploadKeyRequest {
-  string server_id   = 1;
-  string agent_token = 2;
-  string public_key  = 3;
-  string label       = 4;
-}
-message UploadKeyResponse {
-  string key_id = 1;
+  rpc Register(RegisterRequest)              returns (RegisterResponse);
+  rpc SyncKeys(SyncRequest)                  returns (SyncResponse);
+  rpc UploadGeneratedKey(UploadKeyRequest)   returns (UploadKeyResponse);
+  rpc ReportUpdates(ReportUpdatesRequest)    returns (ReportUpdatesResponse);
+  rpc ReportInventory(InventoryReport)       returns (InventoryReportResponse);
+  rpc SyncMonitors(SyncMonitorsRequest)      returns (SyncMonitorsResponse);
+  rpc ReportChecks(ReportChecksRequest)      returns (ReportChecksResponse);
+  rpc CommandStream(stream AgentMessage)     returns (stream ServerCommand);
 }
 ```
 
-No streaming — polling only. Poll interval: **30 seconds**.
+`CommandStream` is the only streaming RPC: the agent authenticates once with `AgentReady`, then the server pushes `ServerCommand`s and the agent replies with `CommandResult`, `StepResult`, or `StepOutputChunk`.
+
+`ServerCommand` variants: `GenerateKeyCmd`, `DeleteKeyCmd`, `UpdateAgentCmd`, `ApplyUpdatesCmd`, `RunStepCmd`, `CleanupWorkspaceCmd`.
+
+Key-state polling stays on the 30s `SyncKeys` interval. Full message definitions live in `proto/vantage/v1/vantage.proto`.
+
+---
+
+## REST API
+
+Unauthenticated:
+
+```
+GET  /install            /install.ps1        # dynamic agent install scripts
+GET  /update             /update.ps1
+GET  /auth/bootstrap-status
+POST /auth/bootstrap     /auth/login  /auth/logout
+GET  /auth/me            /auth/oidc/start    /auth/oidc/callback
+GET  /api/secrets/:group/values               # bearer token (ESO)
+```
+
+Session-authed under `/api`:
+
+```
+servers      GET,POST /servers · GET,POST /servers/new · GET,DELETE /servers/:id
+             POST /servers/:id/{generate-key,update-agent,apply-updates}
+keys         GET,POST /keys · GET,DELETE /keys/:id · GET /keys/:id/private-key
+             POST /keys/:id/assign · DELETE /keys/:id/assign/:serverId
+workflows    GET,POST /steps · PUT,DELETE /steps/:id · GET /steps/:id/export
+             POST /steps/{import,seed-defaults,parse} · GET /steps/usage
+             GET,POST /workflows · GET,PUT,DELETE /workflows/:id
+             POST /workflows/:id/run · GET /workflows/:id/runs
+             GET /runs/:runId · POST /runs/:runId/cancel
+             GET /runs/:runId/servers/:serverId/logs[/stream]
+monitors     GET,POST /monitors · GET,PUT,DELETE /monitors/:id
+             GET /monitors/:id/{incidents,uptime}
+channels     GET,POST /channels · PUT,DELETE /channels/:id · POST /channels/:id/test
+secrets      GET,POST /secrets · GET,PUT,DELETE /secrets/:group
+             POST /secrets/:group/reveal · DELETE /secrets/:group/:key
+console      POST /console/connect · GET /console/tunnel (websocket)
+audit        GET /audit
+agent        GET /agent/latest-version
+settings     GET,PUT /settings · POST /settings/secrets-token       (owner|admin)
+org          GET,POST /org/users · PUT /org/users/:id/role · DELETE /org/users/:id
+             GET,PUT /org/oidc                                       (owner|admin)
+```
 
 ---
 
 ## MongoDB Collections
 
-### `servers`
+`servers` · `keys` · `assignments` · `orgs` · `users` · `org_oidc` · `settings` · `secrets` · `workflows` · `workflow_steps` · `workflow_runs` · `monitors` · `incidents` · `monitor_rollups` · `notification_channels` · `console_sessions` · `audit_logs` · `migrations`
 
-```json
-{
-    "_id": "ObjectId",
-    "server_id": "uuid",
-    "hostname": "proxmox-node-1",
-    "ip_address": "10.10.10.5",
-    "os_info": "Ubuntu 24.04",
-    "pre_reg_token": "abc123",
-    "pre_reg_expires": "ISODate",
-    "agent_token_hash": "sha256...",
-    "status": "pending|active|offline",
-    "last_seen": "ISODate",
-    "created_at": "ISODate"
-}
-```
+Every document except `migrations` carries `org_id`. Struct definitions are the source of truth — see `server/internal/models/`.
 
-- `pre_reg_token` is cleared after the agent successfully calls `Register()`
-- `agent_token_hash` stores SHA-256 of the token — never plaintext
-- `status` transitions: `pending` → `active` on first `Register()`, `offline` if last_seen exceeds threshold
+`site_pending_signups` is written only by sitesvc and holds unverified signups; the control plane neither reads nor knows about it.
 
-### `keys`
+Notes that are not obvious from the structs:
 
-```json
-{
-    "_id": "ObjectId",
-    "key_id": "uuid",
-    "label": "dom-macbook",
-    "public_key": "ssh-ed25519 AAAA...",
-    "fingerprint": "SHA256:...",
-    "source": "uploaded|generated",
-    "generated_by_server_id": "uuid",
-    "created_at": "ISODate"
-}
-```
+- `servers.agent_token_hash` stores SHA-256 of the token, never plaintext. `pre_reg_token` is cleared after `Register()`. `status` is `pending` → `active` on register, `offline` when `last_seen` passes the threshold (swept every 2 min).
+- `servers.inventory` holds the latest metrics snapshot with separate `metrics_at` / `static_at` timestamps.
+- `keys.private_key_enc` and `passphrase_enc` are AES-256-GCM; the JSON form exposes only `has_private_key` / `has_passphrase`.
+- `assignments.revoked_at: null` means active. Revocation is soft, preserving audit history.
+- `workflow_runs.steps_snapshot` freezes the resolved steps so editing the library never rewrites history.
+- `console_sessions.token_consumed_at` is set atomically to enforce one-time use.
 
-### `assignments`
+### Migrations
 
-```json
-{
-    "_id": "ObjectId",
-    "key_id": "uuid",
-    "server_id": "uuid",
-    "assigned_at": "ISODate",
-    "revoked_at": "ISODate | null"
-}
-```
+`services.RunMigrations()` runs at boot, recording markers in `migrations`:
 
-- `revoked_at: null` = key is active on that server
-- Revocation is soft — set `revoked_at`, agent picks it up on next poll
+- `0001_default_org_backfill`
+- `0002_settings_org_backfill` (must run before 0003 — 0003 can create a `default` org, which pushes 0002 into its ambiguous multi-org branch)
+- `0003_missed_org_scopes`
+
+Index builders (`EnsureAuthIndexes`, `EnsureSettingsIndexes`) are fatal on failure; `EnsureSecretIndexes` and `EnsureWorkflowIndexes` only warn.
 
 ---
 
 ## Agent Lifecycle
 
-### Config file — `/etc/vantage/config.yaml`
+### Config file
+
+Linux `/etc/vantage/config.yaml`, Windows `%ProgramData%\vantage\config.yaml`. Directory `0700`, file `0600`.
 
 ```yaml
 server_url: "vantage.yourdomain.com:9090"
 server_id: "<uuid>"
-pre_reg_token: "<token>" # removed after first successful Register()
-agent_token: "" # written by agent after Register()
+pre_reg_token: "<token>"   # removed after first successful Register()
+agent_token: ""            # written by agent after Register()
 poll_interval: 30s
 tls: true
 ```
 
-Config file permissions: `0600`. Config directory: `0700`.
-
-### Startup flow
+### Startup
 
 ```
 1. Load config
-2. If pre_reg_token present:
-     → call Register(server_id, pre_reg_token, hostname, ip, os_info)
-     → save returned agent_token to config
-     → delete pre_reg_token from config
-3. Enter poll loop
+2. If pre_reg_token present → Register() → save agent_token, clear pre_reg_token, reconnect
+3. Start goroutines: command stream · update check (hourly) · inventory · monitors
+4. Enter SyncKeys poll loop (default 30s)
 ```
 
-### Poll loop (every 30s)
+### Poll loop
 
 ```
-1. Call SyncKeys(server_id, agent_token)
-2. Receive []public_keys
-3. Compute fingerprints of current /root/.ssh/authorized_keys
-4. If state unchanged → skip write
-5. If changed:
-     → write to /root/.ssh/authorized_keys.tmp
-     → os.Rename() to /root/.ssh/authorized_keys  (atomic)
-     → chmod 0600
+1. SyncKeys(server_id, agent_token, agent_version)
+2. Non-Linux hosts stop here — Windows agents register and heartbeat only
+3. Diff desired keys against /root/.ssh/authorized_keys; unchanged → no write
+4. Changed → write .tmp, os.Rename() over the real file, chmod 0600
 ```
 
-### Key generation (on demand)
+### Install
 
-- Triggered by a flag or API call from the server
-- Runs `ssh-keygen` via `exec.Command`
-- Uploads public key via `UploadGeneratedKey()`
-- Private key stays local on the machine
-
-### Systemd unit — `/etc/systemd/system/vantage-agent.service`
-
-```ini
-[Unit]
-Description=Vantage Agent
-After=network.target
-
-[Service]
-ExecStart=/usr/local/bin/vantage-agent
-Restart=always
-RestartSec=10
-User=root
-
-[Install]
-WantedBy=multi-user.target
-```
+Linux: systemd unit at `/etc/systemd/system/vantage-agent.service`, `Restart=always`, runs as root.
+Windows: MSI built by CI (WiX), or `installer/setup.ps1` registering the agent as a service via NSSM.
 
 ---
 
 ## Server Registration Flow
 
-1. Click **Add Server** in the UI
-2. Backend generates a short-lived pre-registration token (TTL: 1 hour) and a `server_id`
-3. UI displays a one-liner install command with copy button:
-    ```bash
-    curl -fsSL https://vantage.yourdomain.com/install | \
-      bash -s -- --server-id=<id> --token=<token>
-    ```
-4. Install script:
-    - Detects arch (`amd64` / `arm64`)
-    - Downloads agent binary from Gitea release
-    - Verifies SHA-256 checksum
-    - Writes `/etc/vantage/config.yaml`
-    - Installs and starts systemd unit
-5. On first `SyncKeys` call, server marks status as `active`
+1. **Add Server** in the UI calls `POST /api/servers/new`, which generates a `server_id` and a pre-registration token (TTL 1 hour, single-use).
+2. The UI shows a one-liner:
+   ```bash
+   curl -fsSL https://vantage.yourdomain.com/install | \
+     bash -s -- --server-id=<id> --token=<token>
+   ```
+   Windows gets the `/install.ps1` equivalent.
+3. The script detects arch, downloads the agent from the Gitea release, verifies the SHA-256 checksum, writes the config, installs and starts the service.
+4. The server flips to `active` on first sync.
 
-The backend serves `/install` dynamically, injecting the latest agent version by querying the Gitea API for the most recent `agent/v*` release tag.
+`/install` is served dynamically, injecting the latest agent version from the Gitea API.
+
+---
+
+## Environment Variables (server)
+
+| Name | Required | Notes |
+| --- | --- | --- |
+| `GRPC_HOST` | **yes** | `host:port` agents dial. Boot fails without it — there is no safe default; falling back to the web host would hand agents a port that does not speak gRPC. |
+| `MONGO_URI` | no | default `mongodb://localhost:27017` |
+| `MONGO_DB` | no | default `vantage` |
+| `REDIS_ADDR` | no | default `localhost:6379` |
+| `KEY_ENCRYPTION_KEY` | yes in practice | 64-char hex (32 bytes) for AES-256-GCM. Required for private keys, secrets, OIDC secrets, RDP credentials. |
+| `GITEA_HOST` | yes | used to build install scripts and agent download URLs |
+| `GUACD_ADDR` | no | default `guacd:4822` |
+| `APP_ROOT_LABEL` | no | default `vantage`; wrong value disables the host/session org guard |
+| `VANTAGE_WORKFLOW_LOG_DIR` | no | where run logs are written |
+
+**sitesvc** (`deploy/docker-compose.site.yml` only):
+
+| Name | Required | Notes |
+| --- | --- | --- |
+| `MONGO_URI` | yes | **must point at the control plane's database**, or the app will not see organisations created here. The database name is read from the URI path (`mongodb://user:pass@host:27017/vantage?authSource=vantage`); a URI without one is refused at boot rather than defaulted. Note this differs from the server, which takes `MONGO_DB` separately. |
+| `PUBLIC_URL` | yes | sitesvc's own public base URL; verification links are built from it |
+| `APP_LOGIN_URL` | no | where a verified owner is sent to sign in; without it they get a plain confirmation page |
+| `SMTP_HOST` / `SMTP_FROM` | yes | without them both forms refuse (503) rather than silently dropping |
+| `SMTP_TO` | no | default `support@hostxtra.co.uk`; contact enquiries only |
+| `SMTP_PORT` | no | default `587`; `465` uses implicit TLS |
+| `SMTP_USERNAME` / `SMTP_PASSWORD` | no | auth skipped when username is empty |
+| `SITE_ORIGIN` | yes in practice | comma-separated allowed origins; unset refuses every cross-origin browser request |
+| `TRUST_PROXY` | no | only `true` behind a proxy that overwrites `X-Forwarded-For`, or clients spoof past the rate limiter |
+
+`deploy/docker-compose.yml` runs four services: `redis`, `guacd`, `server` (8080 + 9090), `web` (3000). MongoDB is external. `deploy/docker-compose.site.yml` adds the public marketing site on `3001` and is only used on vantage.sh.
 
 ---
 
 ## Security
 
-- gRPC over TLS (Let's Encrypt or self-signed with cert pinning on the agent)
-- Agent authenticates with a per-server token stored at `/etc/vantage/config.yaml` (`0600`)
-- Server stores `SHA-256(agent_token)` — never the plaintext token
-- Private keys generated by agents are encrypted at rest in MongoDB (AES-256)
-- `authorized_keys` written as `0600`, owned by root
-- Pre-registration tokens are short-lived (1 hour) and single-use
-- Agent runs as `root` (required for `/root/.ssh/authorized_keys` writes)
+- gRPC over TLS; agents connect outbound only, no inbound firewall holes on managed servers.
+- Per-server agent token stored as SHA-256 on the server, plaintext only in the agent's `0600` config.
+- Pre-registration tokens are short-lived (1 hour) and single-use.
+- AES-256-GCM at rest for private keys, key passphrases, vault secrets, OIDC client secrets, RDP/VNC credentials.
+- Console session tokens are one-time; RDP credentials are consumed on tunnel open.
+- ESO read token stored as a SHA-256 hash and rotatable.
+- Unique indexes on user email, org slug, settings org, and ESO token hash are load-bearing for tenant isolation.
+- `authorized_keys` written `0600`, owned by root. The agent runs as root because it must.
+- Every mutating API path writes an audit event.
 
 ---
 
-## Frontend Routes
+## Frontend
 
-| Route           | Purpose                                                              |
-| --------------- | -------------------------------------------------------------------- |
-| `/servers`      | List all servers, online/offline status badge, last seen timestamp   |
-| `/servers/new`  | Displays the one-liner install script with copy button               |
-| `/servers/[id]` | Keys installed on this server, trigger key generation, remove server |
-| `/keys`         | All keys — label, fingerprint, source, assigned count                |
-| `/keys/[id]`    | Assign key to servers, revoke per server                             |
+Next.js 16 (App Router) + React 18, Tailwind 3, TanStack Query. Guacamole client bundled locally in `web/lib/guacamole-common.js`.
+
+| Route | Purpose |
+| --- | --- |
+| `/setup` | First-run bootstrap: create the first org and owner |
+| `/login` | Local or OIDC sign-in |
+| `/` | Fleet dashboard |
+| `/servers`, `/servers/new`, `/servers/[id]` | Fleet list, install one-liner, server detail (keys, inventory, updates) |
+| `/servers/[id]/console` | Browser SSH/RDP/VNC session |
+| `/keys`, `/keys/[id]` | Key library; assign and revoke per server |
+| `/workflows`, `/workflows/[id]`, `/workflows/[id]/runs[/runId]` | Compose, run, and follow live logs |
+| `/steps` | Reusable step library |
+| `/monitors`, `/monitors/new`, `/monitors/[id][/edit]` | Checks, uptime, incidents |
+| `/secrets`, `/secrets/[group]` | Vault |
+| `/audit` | Audit log |
+| `/settings`, `/settings/org`, `/settings/notifications` | Alerts, members, OIDC, channels |
 
 ---
 
 ## CI/CD — Gitea Actions
 
-### Agent release — `.gitea/workflows/agent-release.yml`
+### `agent-release.yml` — triggered by `agent/v*` tags
 
-Triggered by a `agent/v*` tag. Cross-compiles for `linux/amd64` and `linux/arm64`, creates a Gitea release with binaries and checksums.
-
-```yaml
-on:
-    push:
-        tags:
-            - "agent/v*"
-```
-
-Build command:
+Builds `linux/amd64`, `linux/arm64`, `windows/amd64`, writes `checksums.txt`, creates a Gitea release. A second `msi` job on `windows-2022` packages the WiX installer.
 
 ```bash
 GOOS=linux GOARCH=amd64 go build \
@@ -301,51 +400,51 @@ GOOS=linux GOARCH=amd64 go build \
   -o dist/vantage-agent-linux-amd64 ./cmd
 ```
 
-Release assets:
+### `server-deploy.yml` — triggered on every push to `main`
 
-- `vantage-agent-linux-amd64`
-- `vantage-agent-linux-arm64`
-- `checksums.txt`
+Builds and pushes four images to the Gitea container registry: `server`, `web`, `site` and `sitesvc`.
 
-### Server deploy — `.gitea/workflows/server-deploy.yml`
-
-Triggered on pushes to `main` touching `server/**`, `web/**`, or `proto/**`. Builds and pushes Docker images to the Gitea container registry, then deploys via SSH:
+Note that despite the name, **this workflow does not deploy** — it only builds and pushes. There is no SSH step and no path filter; every push to `main` rebuilds all three images. Rolling them out is a separate manual step on the host:
 
 ```bash
-cd /opt/vantage && docker compose pull && docker compose up -d --remove-orphans
+cd /opt/vantage && docker compose -f docker-compose.yml -f docker-compose.site.yml pull && \
+  docker compose -f docker-compose.yml -f docker-compose.site.yml up -d --remove-orphans
 ```
 
-### Tagging convention
+### Tagging
 
 ```bash
-# Release a new agent version
-git tag agent/v1.0.0 && git push origin agent/v1.0.0
-
-# Server + web deploy automatically on push to main
-git push origin main
+git tag agent/v1.0.0 && git push origin agent/v1.0.0   # agent release
+git push origin main                                    # server + web deploy
 ```
 
-### Required Gitea secrets / variables
+### Secrets / variables
 
-| Name                | Type     | Value                                      |
-| ------------------- | -------- | ------------------------------------------ |
-| `RELEASE_TOKEN`     | Secret   | Gitea API token with `write:release` scope |
-| `REGISTRY_USER`     | Secret   | Gitea username                             |
-| `REGISTRY_PASSWORD` | Secret   | Gitea token with `write:packages` scope    |
-| `DEPLOY_HOST`       | Secret   | IP/hostname of the server VM               |
-| `DEPLOY_USER`       | Secret   | SSH user for deploy                        |
-| `DEPLOY_SSH_KEY`    | Secret   | Private key for deploy SSH                 |
-| `GITEA_HOST`        | Variable | `gitea.hostxtra.co.uk`                     |
+| Name | Type | Value |
+| --- | --- | --- |
+| `RELEASE_TOKEN` | Secret | Gitea API token, `write:release` |
+| `REGISTRY_USER` | Secret | Gitea username |
+| `REGISTRY_PASSWORD` | Secret | Gitea token, `write:packages` |
+| `GITEA_HOST` | Variable | `gitea.hostxtra.co.uk` |
+| `DOCKER_HOST` | Variable | registry host used for image tags |
+| `API_URL` | Variable | baked into the `web` image at build time |
+| `SITE_API_URL` | Variable | sitesvc base URL, baked into the `site` image (contact form) |
+| `SITE_CONTACT_EMAIL` | Variable | optional; mailto fallback address |
 
 ---
 
 ## Design Decisions
 
-- **gRPC over REST for agent communication** — strong typing, easy versioning, bi-directional streaming available later if push-based updates are needed
-- **Poll-only, no streaming** — 30s interval is sufficient for a homelab; simplifies agent implementation
-- **Outbound-only agent connections** — no inbound firewall holes required on managed servers
-- **Atomic `authorized_keys` rewrite** — write to `.tmp` then `os.Rename()` prevents partial writes
-- **Fingerprint diffing before write** — avoids unnecessary disk writes on unchanged state
-- **Soft revocation** — `revoked_at` timestamp rather than hard deletes; preserves audit history
-- **root only** — manages `/root/.ssh/authorized_keys` only; no per-user key management
-- **Gitea releases for agent binaries** — slots into existing act_runner CI pipeline; install script queries Gitea API for latest version at serve time
+- **gRPC for agent traffic** — strong typing and cheap versioning; polling for state, one bidirectional stream for commands.
+- **Outbound-only agents** — no inbound ports on managed servers, works behind NAT.
+- **Poll for keys, push for commands** — a 30s key poll is fine, but running a workflow step should not wait up to 30s.
+- **Atomic `authorized_keys` rewrite** — temp file plus `os.Rename()`; a machine that dies mid-write keeps the old file.
+- **Fingerprint diffing before write** — no disk churn on unchanged state.
+- **Soft revocation** — `revoked_at` rather than deletes; preserves audit history.
+- **Run snapshots** — workflow runs freeze their resolved steps so editing a step never rewrites past runs.
+- **Monitors run in two places** — server-side for external endpoints, agent-side for anything only reachable from inside the target network.
+- **Redis for sessions only** — all durable state stays in MongoDB; losing Redis logs everyone out and nothing else.
+- **guacd for console** — protocol handling is Guacamole's problem, not ours; we proxy the WebSocket and manage credentials.
+- **`org_id` on every document** — isolation enforced at the query layer, not by separate databases.
+- **root only** — manages `/root/.ssh/authorized_keys`; no per-user key management.
+- **Windows agents are second-class by design** — register, heartbeat, run steps, report inventory; no `authorized_keys` management.

deploy/docker-compose.site.yml

@@ -0,0 +1,48 @@
+# Public marketing site and its backend. Deliberately kept out of
+# docker-compose.yml so a self-hosted install never runs either of them:
+#
+#   self-hosted:  docker compose up -d
+#   vantage.sh:   docker compose -f docker-compose.yml -f docker-compose.site.yml up -d
+#
+# sitesvc owns both public forms end to end. It shares MongoDB with the control
+# plane — that is how a new tenant becomes visible to the app — but shares no
+# code and no process with it. The control plane has no public signup endpoint.
+services:
+    site:
+        image: gitea.hostxtra.co.uk/mrhid6/vantage/site:latest
+        restart: unless-stopped
+        ports:
+            - 3001:3000
+        depends_on:
+            - sitesvc
+
+    sitesvc:
+        image: gitea.hostxtra.co.uk/mrhid6/vantage/sitesvc:latest
+        restart: unless-stopped
+        ports:
+            - 8082:8082
+        environment:
+            PORT: "8082"
+            # Must point at the same database the control plane uses, or the app
+            # will not see organisations created here. The database name comes
+            # from the URI path:
+            #   mongodb://user:pass@host:27017/vantage?authSource=vantage
+            # A URI with no database is refused at boot rather than defaulted.
+            MONGO_URI: ${MONGO_URI:-}
+            # Public base URL of this service. Verification links are built from
+            # it, so an unset or wrong value produces links that go nowhere.
+            PUBLIC_URL: ${SITE_PUBLIC_URL:-}
+            # Where a verified owner is sent to sign in.
+            APP_LOGIN_URL: ${SITE_APP_LOGIN_URL:-}
+            # Origins allowed to POST the forms. Unset means every cross-origin
+            # browser request is refused.
+            SITE_ORIGIN: ${SITE_ORIGIN:-}
+            # Only enable behind a proxy that overwrites X-Forwarded-For;
+            # otherwise clients can spoof their way past the rate limiter.
+            TRUST_PROXY: ${SITE_TRUST_PROXY:-false}
+            SMTP_HOST: ${SITE_SMTP_HOST:-}
+            SMTP_PORT: ${SITE_SMTP_PORT:-587}
+            SMTP_USERNAME: ${SITE_SMTP_USERNAME:-}
+            SMTP_PASSWORD: ${SITE_SMTP_PASSWORD:-}
+            SMTP_FROM: ${SITE_SMTP_FROM:-}
+            SMTP_TO: ${SITE_SMTP_TO:-support@hostxtra.co.uk}

deploy/docker-compose.yml

@@ -27,19 +27,17 @@ services:
       MONGO_URI: ${MONGO_URI:-}
       REDIS_ADDR: redis:6379
       GITEA_HOST: ${GITEA_HOST}
-      PUBLIC_HOST: ${PUBLIC_HOST}
       GRPC_HOST: ${GRPC_HOST}
       GRPC_PORT: "9090"
       HTTP_PORT: "8080"
-      OIDC_ISSUER: ${OIDC_ISSUER:-}
-      OIDC_CLIENT_ID: ${OIDC_CLIENT_ID:-}
-      OIDC_CLIENT_SECRET: ${OIDC_CLIENT_SECRET:-}
-      OIDC_REDIRECT_URL: ${OIDC_REDIRECT_URL:-}
       KEY_ENCRYPTION_KEY: ${KEY_ENCRYPTION_KEY:-}
+      VANTAGE_WORKFLOW_LOG_DIR: ${VANTAGE_WORKFLOW_LOG_DIR:-}
       GUACD_ADDR: guacd:4822
     depends_on:
       redis:
         condition: service_healthy
+    volumes:
+      - ./data:/data
   web:
     image: gitea.hostxtra.co.uk/mrhid6/vantage/web:latest
     restart: unless-stopped

docs/superpowers/plans/2026-07-20-fleet-inventory.md

@@ -1,653 +0,0 @@
-# Fleet Inventory Implementation Plan
-
-> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
-
-**Goal:** Agents collect CPU/RAM/swap/disk/partition inventory and report it to the server via a new `ReportInventory` RPC; the server stores the latest snapshot per server and the UI displays it.
-
-**Architecture:** New unary gRPC `ReportInventory` (mirrors existing `ReportUpdates`). Agent runs a 30s metrics ticker (CPU/RAM/swap usage) and, every 15 min, a full static collection (disks, partitions, CPU model, kernel). Server upserts an embedded `inventory` sub-doc on the `servers` document with merge rules that preserve static fields between slow ticks.
-
-**Tech Stack:** Go (gin, mongo-driver v2, hand-written JSON-codec gRPC), `/proc` readers, Next.js 16 + react-query + Tailwind.
-
-## Global Constraints
-
-- **No tests this iteration.** Verify with `go build ./...`, `go vet ./...`, `npm run build`.
-- gRPC uses a JSON codec: edit **both** `server/internal/grpc/pb/vantage.pb.go` and `agent/internal/grpc/pb/vantage.pb.go` identically, plus `proto/vantage/v1/vantage.proto` as documentation. No codegen. Mirror the existing `ReportUpdates` RPC wiring exactly (service interface, `_Vantage_*_Handler`, client method, `Vantage_ServiceDesc`).
-- Mongo: `db.Col("servers")`, `context.WithTimeout`. Follow `server/internal/services/servers.go`.
-- Agent already runs as root; `/proc` is readable. Linux is primary; Windows collectors may return empty.
-- Module path `github.com/mrhid6/vantage`.
-- Do not add heavy dependencies; implement `/proc` parsing directly.
-
----
-
-## Task 1: Inventory model + gRPC messages
-
-**Files:**
-- Modify: `server/internal/models/server.go`
-- Modify: `proto/vantage/v1/vantage.proto`
-- Modify: `server/internal/grpc/pb/vantage.pb.go`
-- Modify: `agent/internal/grpc/pb/vantage.pb.go`
-
-**Interfaces:**
-- Produces: `models.Inventory` (+ `CPUInfo`, `MemInfo`, `Partition`) and `Server.Inventory *Inventory`. pb structs `InventoryReport`, `CPUReport`, `MemReport`, `PartitionReport`, `InventoryReportResponse`. Service method `ReportInventory` on both client and server interfaces.
-
-- [ ] **Step 1: Add model structs**
-
-In `server/internal/models/server.go` add (keep the existing `import "time"`):
-
-```go
-type CPUInfo struct {
-	Model    string  `bson:"model,omitempty" json:"model,omitempty"`
-	Cores    int     `bson:"cores,omitempty" json:"cores,omitempty"`
-	UsagePct float64 `bson:"usage_pct" json:"usage_pct"`
-	Load1    float64 `bson:"load1,omitempty" json:"load1,omitempty"`
-}
-
-type MemInfo struct {
-	TotalBytes uint64 `bson:"total_bytes" json:"total_bytes"`
-	UsedBytes  uint64 `bson:"used_bytes" json:"used_bytes"`
-}
-
-type Partition struct {
-	Device     string `bson:"device" json:"device"`
-	Mountpoint string `bson:"mountpoint" json:"mountpoint"`
-	Fstype     string `bson:"fstype,omitempty" json:"fstype,omitempty"`
-	TotalBytes uint64 `bson:"total_bytes" json:"total_bytes"`
-	UsedBytes  uint64 `bson:"used_bytes" json:"used_bytes"`
-}
-
-type Inventory struct {
-	CPU            CPUInfo     `bson:"cpu" json:"cpu"`
-	Memory         MemInfo     `bson:"memory" json:"memory"`
-	SwapTotalBytes uint64      `bson:"swap_total_bytes" json:"swap_total_bytes"`
-	SwapUsedBytes  uint64      `bson:"swap_used_bytes" json:"swap_used_bytes"`
-	Partitions     []Partition `bson:"partitions,omitempty" json:"partitions,omitempty"`
-	Kernel         string      `bson:"kernel,omitempty" json:"kernel,omitempty"`
-	MetricsAt      *time.Time  `bson:"metrics_at,omitempty" json:"metrics_at,omitempty"`
-	StaticAt       *time.Time  `bson:"static_at,omitempty" json:"static_at,omitempty"`
-}
-```
-
-Add to the `Server` struct: `Inventory *Inventory \`bson:"inventory,omitempty" json:"inventory,omitempty"\``.
-
-- [ ] **Step 2: Document RPC in proto**
-
-In `proto/vantage/v1/vantage.proto`, add to the service: `rpc ReportInventory(InventoryReport) returns (InventoryReportResponse);` and the messages `InventoryReport`, `CPUReport`, `MemReport`, `PartitionReport`, `InventoryReportResponse` per spec §4.
-
-- [ ] **Step 3: Add pb structs + RPC wiring (server pb)**
-
-In `server/internal/grpc/pb/vantage.pb.go` add the message structs:
-
-```go
-type CPUReport struct {
-	Model    string  `json:"model,omitempty"`
-	Cores    int     `json:"cores,omitempty"`
-	UsagePct float64 `json:"usage_pct"`
-	Load1    float64 `json:"load1,omitempty"`
-}
-type MemReport struct {
-	TotalBytes uint64 `json:"total_bytes"`
-	UsedBytes  uint64 `json:"used_bytes"`
-}
-type PartitionReport struct {
-	Device     string `json:"device"`
-	Mountpoint string `json:"mountpoint"`
-	Fstype     string `json:"fstype,omitempty"`
-	TotalBytes uint64 `json:"total_bytes"`
-	UsedBytes  uint64 `json:"used_bytes"`
-}
-type InventoryReport struct {
-	ServerId      string            `json:"server_id"`
-	AgentToken    string            `json:"agent_token"`
-	IncludeStatic bool              `json:"include_static"`
-	CPU           *CPUReport        `json:"cpu,omitempty"`
-	Memory        *MemReport        `json:"memory,omitempty"`
-	SwapTotal     uint64            `json:"swap_total"`
-	SwapUsed      uint64            `json:"swap_used"`
-	Partitions    []PartitionReport `json:"partitions,omitempty"`
-	Kernel        string            `json:"kernel,omitempty"`
-}
-type InventoryReportResponse struct{}
-```
-
-Then mirror the `ReportUpdates` RPC plumbing for `ReportInventory`. Locate every `ReportUpdates` reference in this file and add the parallel `ReportInventory`:
-- `VantageServer` interface: add `ReportInventory(context.Context, *InventoryReport) (*InventoryReportResponse, error)`.
-- `UnimplementedVantageServer`: add the stub returning `Unimplemented`.
-- `VantageClient` interface + `keyManagerClient`: add the client method `Invoke`-ing `/vantage.v1.Vantage/ReportInventory`.
-- `Vantage_ServiceDesc.Methods`: add `{MethodName: "ReportInventory", Handler: _Vantage_ReportInventory_Handler}`.
-- Add `_Vantage_ReportInventory_Handler` copied from `_Vantage_ReportUpdates_Handler` with types swapped.
-
-- [ ] **Step 4: Mirror pb structs + wiring (agent pb)**
-
-Apply the identical additions to `agent/internal/grpc/pb/vantage.pb.go`.
-
-- [ ] **Step 5: Verify build**
-
-Run: `cd server && go build ./... && cd ../agent && go build ./...`
-Expected: both succeed.
-
-- [ ] **Step 6: Commit**
-
-```bash
-git add server/internal/models/server.go proto/vantage/v1/vantage.proto server/internal/grpc/pb/vantage.pb.go agent/internal/grpc/pb/vantage.pb.go
-git commit -m "feat(proto): add ReportInventory RPC and inventory model"
-```
-
----
-
-## Task 2: Server handler + store service
-
-**Files:**
-- Create: `server/internal/services/inventory.go`
-- Modify: `server/internal/grpc/server.go`
-
-**Interfaces:**
-- Consumes: `pb.InventoryReport` (T1), `db.Col("servers")`.
-- Produces: `services.StoreInventory(serverID string, r *pb.InventoryReport) error`; gRPC method `(*vantageServer).ReportInventory`.
-
-- [ ] **Step 1: Write the store service**
-
-```go
-package services
-
-import (
-	"context"
-	"time"
-
-	"github.com/mrhid6/vantage/server/internal/db"
-	"github.com/mrhid6/vantage/server/internal/grpc/pb"
-	"go.mongodb.org/mongo-driver/v2/bson"
-)
-
-// StoreInventory upserts the latest inventory snapshot onto the server document.
-// Metrics fields update every call; static fields only when r.IncludeStatic.
-func StoreInventory(serverID string, r *pb.InventoryReport) error {
-	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
-	defer cancel()
-
-	now := time.Now()
-	set := bson.M{"inventory.metrics_at": now}
-	if r.CPU != nil {
-		set["inventory.cpu.usage_pct"] = r.CPU.UsagePct
-		set["inventory.cpu.load1"] = r.CPU.Load1
-	}
-	if r.Memory != nil {
-		set["inventory.memory.used_bytes"] = r.Memory.UsedBytes
-	}
-	set["inventory.swap_used_bytes"] = r.SwapUsed
-
-	if r.IncludeStatic {
-		set["inventory.static_at"] = now
-		set["inventory.swap_total_bytes"] = r.SwapTotal
-		set["inventory.kernel"] = r.Kernel
-		if r.CPU != nil {
-			set["inventory.cpu.model"] = r.CPU.Model
-			set["inventory.cpu.cores"] = r.CPU.Cores
-		}
-		if r.Memory != nil {
-			set["inventory.memory.total_bytes"] = r.Memory.TotalBytes
-		}
-		parts := make([]bson.M, 0, len(r.Partitions))
-		for _, p := range r.Partitions {
-			parts = append(parts, bson.M{
-				"device": p.Device, "mountpoint": p.Mountpoint, "fstype": p.Fstype,
-				"total_bytes": p.TotalBytes, "used_bytes": p.UsedBytes,
-			})
-		}
-		set["inventory.partitions"] = parts
-	}
-
-	_, err := db.Col("servers").UpdateOne(ctx, bson.M{"server_id": serverID}, bson.M{"$set": set})
-	return err
-}
-```
-
-- [ ] **Step 2: Add the gRPC handler**
-
-In `server/internal/grpc/server.go`, add (mirroring the existing `ReportUpdates` handler that validates the agent token):
-
-```go
-func (s *vantageServer) ReportInventory(ctx context.Context, req *pb.InventoryReport) (*pb.InventoryReportResponse, error) {
-	srv, err := services.ValidateAgentToken(req.ServerId, req.AgentToken)
-	if err != nil {
-		return nil, status.Errorf(codes.Unauthenticated, "invalid agent token")
-	}
-	if err := services.StoreInventory(srv.ServerID, req); err != nil {
-		log.Printf("store inventory for %s: %v", srv.ServerID, err)
-	}
-	return &pb.InventoryReportResponse{}, nil
-}
-```
-
-Confirm `status`, `codes`, `log` are already imported in the file (they are, used by other handlers).
-
-- [ ] **Step 3: Verify build**
-
-Run: `cd server && go build ./... && go vet ./...`
-Expected: success.
-
-- [ ] **Step 4: Commit**
-
-```bash
-git add server/internal/services/inventory.go server/internal/grpc/server.go
-git commit -m "feat(server): store inventory and handle ReportInventory RPC"
-```
-
----
-
-## Task 3: Agent collectors
-
-**Files:**
-- Create: `agent/internal/inventory/collect_linux.go`
-- Create: `agent/internal/inventory/collect_other.go`
-- Create: `agent/internal/inventory/inventory.go`
-
-**Interfaces:**
-- Produces: `inventory.Collect(includeStatic bool) *pb.InventoryReport`.
-
-- [ ] **Step 1: Common entry (`inventory.go`)**
-
-```go
-package inventory
-
-import "github.com/mrhid6/vantage/agent/internal/grpc/pb"
-
-// Collect gathers metrics always and static hardware info when includeStatic.
-// Platform specifics are provided by collect_linux.go / collect_other.go.
-func Collect(includeStatic bool) *pb.InventoryReport {
-	r := &pb.InventoryReport{IncludeStatic: includeStatic, CPU: &pb.CPUReport{}, Memory: &pb.MemReport{}}
-	collect(r, includeStatic)
-	return r
-}
-```
-
-- [ ] **Step 2: Linux collector (`collect_linux.go`)**
-
-Build-tagged `//go:build linux`. Implement `collect(r *pb.InventoryReport, includeStatic bool)`:
-- CPU usage: read `/proc/stat` first line twice ~100ms apart, compute `1 - idleDelta/totalDelta` × 100 → `r.CPU.UsagePct`.
-- Load: first field of `/proc/loadavg` → `r.CPU.Load1`.
-- Mem/swap: parse `/proc/meminfo` (`MemTotal`, `MemAvailable`, `SwapTotal`, `SwapFree`; used = total − available; swap used = swaptotal − swapfree) → `r.Memory.*`, `r.SwapUsed`, and on static `r.SwapTotal`.
-- Static only: `/proc/cpuinfo` (`model name`, count `processor` lines) → `r.CPU.Model/Cores`; `/proc/meminfo MemTotal` → `r.Memory.TotalBytes`; kernel via `syscall.Uname` or read `/proc/sys/kernel/osrelease` → `r.Kernel`; partitions from `/proc/mounts` filtered to fstypes in {ext4,xfs,btrfs,zfs,vfat,ntfs} then `syscall.Statfs` for total/used → `r.Partitions`.
-
-```go
-//go:build linux
-
-package inventory
-
-import (
-	"bufio"
-	"os"
-	"strconv"
-	"strings"
-	"syscall"
-	"time"
-
-	"github.com/mrhid6/vantage/agent/internal/grpc/pb"
-)
-
-func collect(r *pb.InventoryReport, includeStatic bool) {
-	r.CPU.UsagePct = cpuUsage()
-	r.CPU.Load1 = load1()
-	memTotal, memAvail, swapTotal, swapFree := meminfo()
-	if memTotal > memAvail {
-		r.Memory.UsedBytes = memTotal - memAvail
-	}
-	if swapTotal > swapFree {
-		r.SwapUsed = swapTotal - swapFree
-	}
-	if includeStatic {
-		r.Memory.TotalBytes = memTotal
-		r.SwapTotal = swapTotal
-		r.CPU.Model, r.CPU.Cores = cpuStatic()
-		r.Kernel = kernel()
-		r.Partitions = partitions()
-	}
-}
-
-func readProc(path string) string { b, _ := os.ReadFile(path); return string(b) }
-
-func cpuSample() (idle, total uint64) {
-	f, err := os.Open("/proc/stat")
-	if err != nil {
-		return
-	}
-	defer f.Close()
-	sc := bufio.NewScanner(f)
-	if sc.Scan() {
-		fields := strings.Fields(sc.Text()) // cpu user nice system idle iowait ...
-		for i, v := range fields[1:] {
-			n, _ := strconv.ParseUint(v, 10, 64)
-			total += n
-			if i == 3 { // idle
-				idle = n
-			}
-		}
-	}
-	return
-}
-
-func cpuUsage() float64 {
-	i1, t1 := cpuSample()
-	time.Sleep(100 * time.Millisecond)
-	i2, t2 := cpuSample()
-	dt := float64(t2 - t1)
-	if dt <= 0 {
-		return 0
-	}
-	return (1 - float64(i2-i1)/dt) * 100
-}
-
-func load1() float64 {
-	fields := strings.Fields(readProc("/proc/loadavg"))
-	if len(fields) > 0 {
-		v, _ := strconv.ParseFloat(fields[0], 64)
-		return v
-	}
-	return 0
-}
-
-func meminfo() (total, avail, swapTotal, swapFree uint64) {
-	f, err := os.Open("/proc/meminfo")
-	if err != nil {
-		return
-	}
-	defer f.Close()
-	sc := bufio.NewScanner(f)
-	for sc.Scan() {
-		fields := strings.Fields(sc.Text())
-		if len(fields) < 2 {
-			continue
-		}
-		kb, _ := strconv.ParseUint(fields[1], 10, 64)
-		b := kb * 1024
-		switch strings.TrimSuffix(fields[0], ":") {
-		case "MemTotal":
-			total = b
-		case "MemAvailable":
-			avail = b
-		case "SwapTotal":
-			swapTotal = b
-		case "SwapFree":
-			swapFree = b
-		}
-	}
-	return
-}
-
-func cpuStatic() (model string, cores int) {
-	f, err := os.Open("/proc/cpuinfo")
-	if err != nil {
-		return
-	}
-	defer f.Close()
-	sc := bufio.NewScanner(f)
-	for sc.Scan() {
-		line := sc.Text()
-		if strings.HasPrefix(line, "processor") {
-			cores++
-		} else if strings.HasPrefix(line, "model name") && model == "" {
-			if i := strings.Index(line, ":"); i >= 0 {
-				model = strings.TrimSpace(line[i+1:])
-			}
-		}
-	}
-	return
-}
-
-func kernel() string {
-	return strings.TrimSpace(readProc("/proc/sys/kernel/osrelease"))
-}
-
-func partitions() []pb.PartitionReport {
-	allowed := map[string]bool{"ext4": true, "xfs": true, "btrfs": true, "zfs": true, "vfat": true, "ntfs": true, "ext3": true}
-	f, err := os.Open("/proc/mounts")
-	if err != nil {
-		return nil
-	}
-	defer f.Close()
-	var out []pb.PartitionReport
-	seen := map[string]bool{}
-	sc := bufio.NewScanner(f)
-	for sc.Scan() {
-		fields := strings.Fields(sc.Text())
-		if len(fields) < 3 || !allowed[fields[2]] || seen[fields[1]] {
-			continue
-		}
-		seen[fields[1]] = true
-		var st syscall.Statfs_t
-		if syscall.Statfs(fields[1], &st) != nil {
-			continue
-		}
-		total := st.Blocks * uint64(st.Bsize)
-		free := st.Bavail * uint64(st.Bsize)
-		out = append(out, pb.PartitionReport{
-			Device: fields[0], Mountpoint: fields[1], Fstype: fields[2],
-			TotalBytes: total, UsedBytes: total - free,
-		})
-	}
-	return out
-}
-```
-
-- [ ] **Step 3: Non-linux stub (`collect_other.go`)**
-
-```go
-//go:build !linux
-
-package inventory
-
-import "github.com/mrhid6/vantage/agent/internal/grpc/pb"
-
-// collect is a no-op best-effort stub on non-Linux platforms.
-func collect(r *pb.InventoryReport, includeStatic bool) {}
-```
-
-- [ ] **Step 4: Verify build**
-
-Run: `cd agent && go build ./... && go vet ./...`
-Expected: success (build both native and, if convenient, `GOOS=windows go build ./...`).
-
-- [ ] **Step 5: Commit**
-
-```bash
-git add agent/internal/inventory/
-git commit -m "feat(agent): /proc-based inventory collectors"
-```
-
----
-
-## Task 4: Agent client method + scheduler
-
-**Files:**
-- Modify: `agent/internal/grpc/client.go`
-- Modify: the agent main loop (`agent/cmd/main.go` or `agent/internal/sync/sync.go` — wherever the poll loop/tickers live).
-
-**Interfaces:**
-- Consumes: `inventory.Collect` (T3), pb (T1).
-- Produces: `(*Client).ReportInventory(report *pb.InventoryReport) error`; a running ticker that reports metrics every 30s and static every 15 min.
-
-- [ ] **Step 1: Add client method**
-
-In `agent/internal/grpc/client.go`, mirroring `ReportUpdates`:
-
-```go
-func (c *Client) ReportInventory(report *pb.InventoryReport) error {
-	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
-	defer cancel()
-	_, err := c.client.ReportInventory(ctx, report)
-	return err
-}
-```
-
-The report already carries `ServerId`/`AgentToken`; ensure the caller sets them (see Step 2).
-
-- [ ] **Step 2: Add the scheduler to the agent loop**
-
-Find where the agent starts its poll loop (the goroutine that calls `SyncKeys`/`ReportUpdates`). Add a parallel inventory ticker. `serverID`, `agentToken`, and the `*Client` are in scope there:
-
-```go
-	go func() {
-		tick := 0
-		t := time.NewTicker(30 * time.Second)
-		defer t.Stop()
-		report := func(static bool) {
-			r := inventory.Collect(static)
-			r.ServerId = serverID
-			r.AgentToken = agentToken
-			if err := client.ReportInventory(r); err != nil {
-				log.Printf("report inventory: %v", err)
-			}
-		}
-		report(true) // send a full snapshot on startup
-		for range t.C {
-			tick++
-			report(tick%30 == 0) // every 30th tick = 15 min → include static
-		}
-	}()
-```
-
-Add imports `"github.com/mrhid6/vantage/agent/internal/inventory"`, `time`, `log` if missing. Match variable names to the actual loop (e.g. the client may be named `c`).
-
-- [ ] **Step 3: Verify build**
-
-Run: `cd agent && go build ./... && go vet ./...`
-Expected: success.
-
-- [ ] **Step 4: Commit**
-
-```bash
-git add agent/internal/grpc/client.go agent/
-git commit -m "feat(agent): schedule inventory reporting (30s metrics, 15m static)"
-```
-
----
-
-## Task 5: Frontend — inventory panel on server detail
-
-**Files:**
-- Modify: `web/lib/api.ts` (extend the `Server`/server-detail type with `inventory`)
-- Modify: `web/app/servers/[id]/page.tsx` (add panel; enable polling)
-
-**Interfaces:**
-- Consumes: server-detail query.
-
-- [ ] **Step 1: Add the inventory type**
-
-In `web/lib/api.ts`, add and attach to the server type used by the detail page:
-
-```ts
-export interface Inventory {
-  cpu: { model?: string; cores?: number; usage_pct: number; load1?: number };
-  memory: { total_bytes: number; used_bytes: number };
-  swap_total_bytes: number;
-  swap_used_bytes: number;
-  partitions?: { device: string; mountpoint: string; fstype?: string; total_bytes: number; used_bytes: number }[];
-  kernel?: string;
-  metrics_at?: string;
-  static_at?: string;
-}
-```
-
-Add `inventory?: Inventory;` to the server detail interface.
-
-- [ ] **Step 2: Add a `formatBytes` helper + Inventory panel**
-
-In `web/app/servers/[id]/page.tsx`, add a helper and a panel component. Enable polling on the server-detail `useQuery` with `refetchInterval: 30000`.
-
-```tsx
-function formatBytes(n: number): string {
-  if (!n) return "0 B";
-  const u = ["B", "KB", "MB", "GB", "TB"];
-  const i = Math.floor(Math.log(n) / Math.log(1024));
-  return `${(n / Math.pow(1024, i)).toFixed(1)} ${u[i]}`;
-}
-
-function UsageBar({ used, total }: { used: number; total: number }) {
-  const pct = total > 0 ? Math.min(100, (used / total) * 100) : 0;
-  return (
-    <div className="h-2 w-full overflow-hidden rounded-full bg-surface-2">
-      <div className={`h-full rounded-full ${pct > 90 ? "bg-danger" : "bg-accent"}`} style={{ width: `${pct}%` }} />
-    </div>
-  );
-}
-
-function InventoryPanel({ inv }: { inv: Inventory }) {
-  return (
-    <Card>
-      <h2 className="mb-4 text-lg font-semibold text-text-primary">Inventory</h2>
-      <div className="grid gap-4 sm:grid-cols-2">
-        <div>
-          <div className="mb-1 flex justify-between text-sm"><span className="text-text-secondary">CPU</span><span className="text-text-primary">{inv.cpu.usage_pct.toFixed(0)}%</span></div>
-          <UsageBar used={inv.cpu.usage_pct} total={100} />
-          <p className="mt-1 text-xs text-text-secondary">{inv.cpu.model} · {inv.cpu.cores} cores · load {inv.cpu.load1?.toFixed(2)}</p>
-        </div>
-        <div>
-          <div className="mb-1 flex justify-between text-sm"><span className="text-text-secondary">Memory</span><span className="text-text-primary">{formatBytes(inv.memory.used_bytes)} / {formatBytes(inv.memory.total_bytes)}</span></div>
-          <UsageBar used={inv.memory.used_bytes} total={inv.memory.total_bytes} />
-          <div className="mb-1 mt-3 flex justify-between text-sm"><span className="text-text-secondary">Swap</span><span className="text-text-primary">{formatBytes(inv.swap_used_bytes)} / {formatBytes(inv.swap_total_bytes)}</span></div>
-          <UsageBar used={inv.swap_used_bytes} total={inv.swap_total_bytes} />
-        </div>
-      </div>
-      {inv.partitions && inv.partitions.length > 0 && (
-        <div className="mt-5">
-          <h3 className="mb-2 text-sm font-medium text-text-secondary">Partitions</h3>
-          <div className="space-y-3">
-            {inv.partitions.map((p) => (
-              <div key={p.mountpoint}>
-                <div className="mb-1 flex justify-between text-xs">
-                  <span className="font-mono text-text-primary">{p.mountpoint}</span>
-                  <span className="text-text-secondary">{formatBytes(p.used_bytes)} / {formatBytes(p.total_bytes)} · {p.fstype}</span>
-                </div>
-                <UsageBar used={p.used_bytes} total={p.total_bytes} />
-              </div>
-            ))}
-          </div>
-        </div>
-      )}
-      {inv.kernel && <p className="mt-4 text-xs text-text-secondary">Kernel {inv.kernel}</p>}
-    </Card>
-  );
-}
-```
-
-Render `{server.inventory && <InventoryPanel inv={server.inventory} />}` in the page body (ensure `Card`, `Inventory` are imported). Match how the page currently reads the server object.
-
-- [ ] **Step 3: Verify build**
-
-Run: `cd web && npm run build`
-Expected: success.
-
-- [ ] **Step 4: Commit**
-
-```bash
-git add web/lib/api.ts web/app/servers/[id]/page.tsx
-git commit -m "feat(web): inventory panel on server detail"
-```
-
----
-
-## Task 6: End-to-end manual verification
-
-- [ ] **Step 1: Build all**
-
-Run: `cd server && go build ./... && cd ../agent && go build ./... && cd ../web && npm run build`
-Expected: all succeed.
-
-- [ ] **Step 2: Smoke (if environment available)**
-
-With server + Mongo + a connected Linux agent: within ~30s the server detail page shows CPU %, RAM/swap bars; within 15 min (or on agent restart, which sends a full snapshot immediately) partitions, CPU model and kernel appear. Confirm metrics update roughly every 30s.
-
-- [ ] **Step 3: Commit any fixes**
-
-```bash
-git add -A
-git commit -m "fix: fleet inventory verification fixes"
-```
-
----
-
-## Self-Review Notes
-
-- **Spec coverage:** §3 model → T1; §4 RPC → T1; §5 collectors + scheduler → T3, T4; §6 handler/store → T2; §7 frontend → T5. Split cadence (30s metrics / 15m static) in T4 scheduler; merge rules preserving static in T2 `StoreInventory`. Tests omitted per Global Constraints.
-- **Startup snapshot:** agent sends `Collect(true)` immediately so static fields populate without waiting 15 min.
-- **Types consistent:** `InventoryReport` field names identical across proto, both pb files, store service, and TS interface (`usage_pct`, `used_bytes`, `total_bytes`, `swap_*`).
-- **Follow-ups (out of scope):** time-series history, usage alerting, Windows collectors, servers-list CPU/RAM badges.

docs/superpowers/specs/2026-07-20-fleet-inventory-design.md

@@ -1,142 +0,0 @@
-# Fleet Inventory — Design
-
-**Date:** 2026-07-20
-**Status:** Approved (design) — ready for implementation planning
-**Scope:** Fleet Inventory only. Server Workflows and SaaS/auth are separate sub-projects.
-
----
-
-## 1. Summary
-
-Each agent collects hardware/OS inventory about its host and reports it to the server, which stores the latest snapshot per server and surfaces it in the UI. Two cadences:
-
-- **Metrics (near-real-time):** CPU load/usage, RAM used/total, swap used/total — every **30s** (aligned with existing poll rhythm).
-- **Static inventory (slow):** disks, partitions and their usage, CPU model/cores, total RAM, OS details — every **15 min**.
-
-Transport: a **new unary gRPC `ReportInventory` RPC** (mirrors the existing `ReportUpdates` pattern). No streaming.
-
----
-
-## 2. Locked decisions
-
-| Topic | Decision |
-|-------|----------|
-| Transport | New `ReportInventory` unary RPC. |
-| Cadence | Metrics every 30s; static inventory every 15 min. One RPC carries both, but static fields are only populated on the 15-min tick (empty/omitted otherwise → server keeps prior static snapshot). |
-| Storage | Latest snapshot embedded on the `servers` document (`inventory` sub-doc). No history/time-series in v1. |
-| Collection | Pure-Go where practical (`/proc`, `gopsutil`-style). Agent already runs as root. |
-| Platform | Linux primary; Windows agent populates what it can, leaves the rest empty. |
-
----
-
-## 3. Data model
-
-Add an `Inventory` sub-document to the existing `Server` model (`server/internal/models/server.go`):
-
-```go
-type CPUInfo struct {
-	Model    string  `bson:"model,omitempty" json:"model,omitempty"`
-	Cores    int     `bson:"cores,omitempty" json:"cores,omitempty"`
-	UsagePct float64 `bson:"usage_pct" json:"usage_pct"`         // metrics tick
-	Load1    float64 `bson:"load1,omitempty" json:"load1,omitempty"`
-}
-
-type MemInfo struct {
-	TotalBytes uint64 `bson:"total_bytes" json:"total_bytes"`
-	UsedBytes  uint64 `bson:"used_bytes" json:"used_bytes"`       // metrics tick
-}
-
-type Partition struct {
-	Device     string `bson:"device" json:"device"`
-	Mountpoint string `bson:"mountpoint" json:"mountpoint"`
-	Fstype     string `bson:"fstype,omitempty" json:"fstype,omitempty"`
-	TotalBytes uint64 `bson:"total_bytes" json:"total_bytes"`
-	UsedBytes  uint64 `bson:"used_bytes" json:"used_bytes"`
-}
-
-type Inventory struct {
-	CPU            CPUInfo     `bson:"cpu" json:"cpu"`
-	Memory         MemInfo     `bson:"memory" json:"memory"`
-	SwapTotalBytes uint64      `bson:"swap_total_bytes" json:"swap_total_bytes"`
-	SwapUsedBytes  uint64      `bson:"swap_used_bytes" json:"swap_used_bytes"`
-	Partitions     []Partition `bson:"partitions,omitempty" json:"partitions,omitempty"`
-	Kernel         string      `bson:"kernel,omitempty" json:"kernel,omitempty"`
-	MetricsAt      *time.Time  `bson:"metrics_at,omitempty" json:"metrics_at,omitempty"`
-	StaticAt       *time.Time  `bson:"static_at,omitempty" json:"static_at,omitempty"`
-}
-```
-
-Add `Inventory *Inventory` field to `Server`.
-
-Server-side update rules:
-- Metrics fields (`cpu.usage_pct`, `cpu.load1`, `memory.used_bytes`, swap used) always updated + `metrics_at`.
-- Static fields (`cpu.model/cores`, `memory.total_bytes`, `partitions`, `kernel`, swap total) updated only when the report includes them (non-zero/non-empty) + `static_at`.
-
----
-
-## 4. gRPC protocol (`proto/vantage/v1/vantage.proto` + both `pb.go` files)
-
-```protobuf
-rpc ReportInventory(InventoryReport) returns (InventoryReportResponse);
-
-message InventoryReport {
-  string server_id    = 1;
-  string agent_token  = 2;
-  bool   include_static = 3; // true on the 15-min tick
-  CPUReport cpu       = 4;
-  MemReport memory    = 5;
-  uint64 swap_total   = 6;
-  uint64 swap_used    = 7;
-  repeated PartitionReport partitions = 8; // only when include_static
-  string kernel       = 9;                 // only when include_static
-}
-message CPUReport { string model = 1; int32 cores = 2; double usage_pct = 3; double load1 = 4; }
-message MemReport { uint64 total_bytes = 1; uint64 used_bytes = 2; }
-message PartitionReport { string device = 1; string mountpoint = 2; string fstype = 3; uint64 total_bytes = 4; uint64 used_bytes = 5; }
-message InventoryReportResponse {}
-```
-
-Hand-written JSON-codec structs added to `server/internal/grpc/pb/vantage.pb.go` and `agent/internal/grpc/pb/vantage.pb.go`, plus the RPC method wiring (service interface, client method, handler registration) mirroring `ReportUpdates`.
-
----
-
-## 5. Agent collection (`agent/internal/inventory/`)
-
-- `Collect(includeStatic bool) *pb.InventoryReport` — reads:
-  - CPU usage: sample `/proc/stat` delta; load from `/proc/loadavg`; model/cores from `/proc/cpuinfo` (static).
-  - Memory/swap: `/proc/meminfo`.
-  - Partitions: `/proc/mounts` filtered to real filesystems + `statfs` for total/used (static).
-  - Kernel: `uname` / `/proc/version` (static).
-  - Windows: best-effort via `wmic`/PS or leave empty.
-- Scheduler in the agent main loop: a 30s ticker calls `Collect(false)` and `ReportInventory`; every 30th tick (15 min) calls `Collect(true)`.
-- Reuse existing gRPC client; add `Client.ReportInventory(...)` like `ReportUpdates`.
-
-Prefer implementing the `/proc` readers directly (no new heavy deps) unless a `gopsutil` dependency is already vendored.
-
----
-
-## 6. Server handler + service
-
-- gRPC handler `ReportInventory` in `server/internal/grpc/server.go`: validate agent token (`ValidateAgentToken`), then call `services.StoreInventory(serverID, report)`.
-- `services.StoreInventory` (in `server/internal/services/inventory.go`): builds the `$set` per the update rules in §3 and `UpdateOne` on `servers`.
-
----
-
-## 7. Frontend
-
-Surface inventory on the existing server detail page (`web/app/servers/[id]/page.tsx`) — add an "Inventory" panel:
-- CPU usage gauge + model/cores, load.
-- RAM used/total bar, swap bar.
-- Partitions table: device, mount, fstype, used/total with a usage bar.
-- "Updated Xs ago" from `metrics_at`/`static_at`.
-
-Optionally add compact CPU/RAM badges to the servers list (`web/app/servers/page.tsx`). Reuse `@/components/ui` + Tailwind tokens. Poll the server detail query while the page is open (react-query `refetchInterval` ~30s) so metrics stay fresh.
-
----
-
-## 8. Out of scope
-
-- Time-series history / graphs (only latest snapshot stored).
-- Alerting thresholds on usage (settings/alerts is a separate concern).
-- Per-process / network / GPU inventory.
-- Tests (skipped, consistent with the Workflows iteration).

docs/superpowers/specs/2026-07-20-saas-auth-orgs-design.md

@@ -1,142 +0,0 @@
-# SaaS: Auth + Organizations — Design
-
-**Date:** 2026-07-20
-**Status:** Approved (design) — ready for implementation planning
-**Scope:** Local auth + organizations + per-org OIDC, and org-scoping of existing data. Billing/plan-limits explicitly deferred. Fleet Inventory and Server Workflows are separate sub-projects.
-
----
-
-## 1. Summary
-
-Turn Vantage from a single-admin, single global-OIDC tool into a multi-tenant app:
-
-1. **Replace** the global Authentik/env-based OIDC with **local email/password accounts** as the primary login.
-2. **Organizations** — every user belongs to an org; every domain object (servers, keys, secrets, assignments, workflows, steps, runs, audit) carries an `org_id` and all queries are scoped to the caller's org.
-3. **Per-org OpenID** — an org admin can configure their own OIDC provider (issuer/client id/secret); users in that org can then sign in through it.
-
-No billing, no seat/server limits this iteration (schema leaves room).
-
----
-
-## 2. Locked decisions
-
-| Topic | Decision |
-|-------|----------|
-| Primary auth | Local email + password (bcrypt). Replaces global Authentik. |
-| Org SSO | Per-org OIDC provider, configured by org admin, resolved dynamically at login. |
-| Isolation | `org_id` on every collection; every service query filtered by org. Enforced in the request layer via session→org. |
-| Roles | `owner`, `admin`, `member` (v1: owner/admin can manage users + org OIDC + all resources; member can use resources). Keep minimal. |
-| Bootstrapping | First-run creates the initial org + owner account (setup flow) when no users exist. |
-| Sessions | Keep existing Redis session store; session now carries `user_id`, `org_id`, `role`, `email`. |
-| Agent auth | Unchanged (per-server agent tokens). Servers gain `org_id`; agent RPCs resolve org from the server record. |
-
----
-
-## 3. Data model
-
-### `orgs`
-```json
-{ "_id":"ObjectId", "org_id":"uuid", "name":"Acme", "created_at":"ISODate" }
-```
-
-### `users`
-```json
-{
-  "_id":"ObjectId", "user_id":"uuid", "org_id":"uuid",
-  "email":"a@b.com", "password_hash":"bcrypt...", "role":"owner|admin|member",
-  "auth_source":"local|oidc", "created_at":"ISODate", "last_login":"ISODate|null"
-}
-```
-Unique index on `email` (global — email identifies the account and its org).
-
-### `org_oidc` (per-org provider config)
-```json
-{
-  "_id":"ObjectId", "org_id":"uuid",
-  "issuer":"https://id.acme.com", "client_id":"...",
-  "client_secret_enc":"AES...",         // encrypted with existing crypto.go
-  "redirect_url":"https://vantage.../auth/oidc/callback",
-  "enabled": true, "updated_at":"ISODate"
-}
-```
-
-### Existing collections — add `org_id`
-`servers`, `keys`, `assignments`, `secrets`, `workflows`, `workflow_steps`, `workflow_runs`, `audit` each gain `org_id string`. A **migration** backfills all existing documents into a default org (see §7).
-
----
-
-## 4. Auth flows
-
-### Local
-- `POST /auth/register` — only allowed during first-run bootstrap (creates org + owner) OR by an org admin inviting a user (see below). Not open self-serve.
-- `POST /auth/login` — email + password → verify bcrypt → create session with `{user_id, org_id, role, email}`.
-- `POST /auth/logout` — destroy session.
-- `GET /auth/me` — returns current user + org.
-
-### Org-admin user management
-- `GET /api/org/users` / `POST /api/org/users` (create local user in caller's org) / `PUT /api/org/users/:id/role` / `DELETE /api/org/users/:id`.
-
-### Per-org OIDC
-- `GET/PUT /api/org/oidc` — read/save the caller org's provider config (admin only). Secret stored encrypted.
-- `GET /auth/oidc/start?org=<org_id or slug>` — look up org's `org_oidc`, build the OIDC provider on demand (cache per org), redirect to authorize.
-- `GET /auth/oidc/callback` — exchange code, match/provision the user by email **within that org**, create session.
-  - If the email exists in the org → log in. If not → provision a `member` with `auth_source=oidc` (org admin can promote). Reject if email belongs to a different org.
-
-### First-run bootstrap
-- `GET /auth/bootstrap-status` → `{ needs_setup: bool }` (true when `users` is empty).
-- Setup page collects org name + owner email/password → creates org + owner → session.
-
----
-
-## 5. Request scoping
-
-- `auth.Middleware` already loads the session; extend `Session` to include `OrgID`, `UserID`, `Role`. Add helper `auth.OrgID(c) string`.
-- **Every service function that reads/writes a scoped collection takes an `orgID` argument** and adds `"org_id": orgID` to its filter and on insert. Handlers pass `auth.OrgID(c)`.
-- Add a `requireRole(role)` gin middleware for admin-only routes (org user mgmt, org OIDC).
-- Agent-facing gRPC: resolve `org_id` from the `servers` record (already tied to `server_id`); inventory/keys/sync operate on that org implicitly.
-
----
-
-## 6. Removing global Authentik
-
-- Delete/retire env-driven `InitOIDC` global provider (`OIDC_ISSUER` etc.). Keep the `go-oidc`/`oauth2` machinery but move it behind the per-org resolver.
-- `authEnabled` global replaced by "auth always on" (there is always local auth). Update `middleware.go` accordingly (no more `if !authEnabled { next }` bypass — except the bootstrap endpoints and login/register which are unauthenticated).
-- Login page (`web/app/login` or existing) offers: email/password form + "Sign in with your organization's SSO" (enter org, redirect to `/auth/oidc/start`).
-
----
-
-## 7. Migration
-
-One-shot migration run at startup (idempotent):
-1. If `orgs` is empty AND `servers`/`keys`/etc. contain documents without `org_id`: create a **default org** ("Default").
-2. Set `org_id = <default>` on all existing `servers`, `keys`, `assignments`, `secrets`, `workflows`, `workflow_steps`, `workflow_runs`, `audit` documents missing it.
-3. If `OIDC_ISSUER` env was set previously and an admin email is known, optionally seed an owner user (documented manual step) — otherwise first-run bootstrap handles owner creation.
-Guard with a marker (e.g. a `migrations` collection entry) so it runs once.
-
----
-
-## 8. Frontend
-
-- **Login/Setup:** `web/app/login/page.tsx` (email/password + org SSO entry) and `web/app/setup/page.tsx` (first-run). Redirect logic based on `bootstrap-status` and `auth/me`.
-- **Org settings:** `web/app/settings/org/` — members list + invite/create user + role management; OIDC provider form (issuer/client id/secret/enabled).
-- Existing pages unchanged functionally but now implicitly org-scoped by the backend. Show current org + user in the sidebar/header.
-
----
-
-## 9. Security
-
-- Passwords: bcrypt (cost ≥ 12). Never returned.
-- Org OIDC client secret encrypted at rest (reuse `services/crypto.go` AES).
-- Cross-org access prevented at the service layer (org_id in every filter) — the primary isolation boundary. Handlers must never accept an `org_id` from the client; always derive from session.
-- OIDC callback must bind the returned identity to the org that initiated the flow (state carries org_id) to prevent org-mixing.
-- Role checks on all org-admin mutations.
-
----
-
-## 10. Out of scope
-
-- Billing, plans, seat/server limits.
-- Cross-org resource sharing, org switching for a single user (one user = one org in v1).
-- SCIM / directory sync, SAML.
-- Email delivery for invites (create-user sets a password or invite token; email sending deferred — document as manual/console output).
-- Tests (skipped, consistent with prior iterations).

proto/vantage/v1/vantage.proto

@@ -9,6 +9,9 @@ service Vantage {
   rpc SyncKeys(SyncRequest)                    returns (SyncResponse);
   rpc UploadGeneratedKey(UploadKeyRequest)      returns (UploadKeyResponse);
   rpc ReportUpdates(ReportUpdatesRequest)       returns (ReportUpdatesResponse);
+  rpc ReportInventory(InventoryReport)          returns (InventoryReportResponse);
+  rpc SyncMonitors(SyncMonitorsRequest)         returns (SyncMonitorsResponse);
+  rpc ReportChecks(ReportChecksRequest)         returns (ReportChecksResponse);
   // Bidirectional stream: agent sends auth once, server pushes commands.
   rpc CommandStream(stream AgentMessage)        returns (stream ServerCommand);
 }
@@ -82,6 +85,80 @@ message ReportUpdatesRequest {
 
 message ReportUpdatesResponse {}
 
+message CPUReport {
+  string model     = 1;
+  int32  cores     = 2;
+  double usage_pct = 3;
+  double load1     = 4;
+}
+
+message MemReport {
+  uint64 total_bytes = 1;
+  uint64 used_bytes  = 2;
+}
+
+message PartitionReport {
+  string device      = 1;
+  string mountpoint  = 2;
+  string fstype      = 3;
+  uint64 total_bytes = 4;
+  uint64 used_bytes  = 5;
+}
+
+message InventoryReport {
+  string server_id      = 1;
+  string agent_token    = 2;
+  bool   include_static = 3;
+  CPUReport cpu         = 4;
+  MemReport memory      = 5;
+  uint64 swap_total     = 6;
+  uint64 swap_used      = 7;
+  repeated PartitionReport partitions = 8;
+  string kernel         = 9;
+}
+
+message InventoryReportResponse {}
+
+message MonitorSpec {
+  string monitor_id      = 1;
+  string type            = 2;
+  string url             = 3;
+  string host            = 4;
+  int32  port            = 5;
+  string method          = 6;
+  int32  expected_status = 7;
+  string keyword         = 8;
+  int32  tls_warn_days   = 9;
+  int32  interval_sec    = 10;
+  int32  retries         = 11;
+  bool   insecure        = 12;
+}
+
+message SyncMonitorsRequest {
+  string server_id   = 1;
+  string agent_token = 2;
+}
+
+message SyncMonitorsResponse {
+  repeated MonitorSpec monitors = 1;
+}
+
+message CheckResult {
+  string monitor_id       = 1;
+  bool   up               = 2;
+  int32  latency_ms       = 3;
+  string message          = 4;
+  int64  cert_expiry_unix = 5;
+}
+
+message ReportChecksRequest {
+  string server_id   = 1;
+  string agent_token = 2;
+  repeated CheckResult results = 3;
+}
+
+message ReportChecksResponse {}
+
 message ApplyUpdatesCmd {}
 
 message ServerCommand {

server/cmd/main.go

@@ -11,6 +11,7 @@ import (
 	"github.com/mrhid6/vantage/server/internal/auth"
 	"github.com/mrhid6/vantage/server/internal/db"
 	grpcserver "github.com/mrhid6/vantage/server/internal/grpc"
+	"github.com/mrhid6/vantage/server/internal/monitorsched"
 	"github.com/mrhid6/vantage/server/internal/services"
 )
 
@@ -18,19 +19,65 @@ func main() {
 	mongoURI := getEnv("MONGO_URI", "mongodb://localhost:27017")
 	dbName := getEnv("MONGO_DB", "vantage")
 
+	// Agents dial gRPC directly, so there is no sane default: falling back to the
+	// public web host would hand every new agent a config pointing at a port that
+	// does not speak gRPC. Fail loudly at boot instead of at install time.
+	if os.Getenv("GRPC_HOST") == "" {
+		log.Fatal("GRPC_HOST is required (host:port agents dial for gRPC)")
+	}
+
 	if err := db.Connect(mongoURI, dbName); err != nil {
 		log.Fatalf("failed to connect to MongoDB: %v", err)
 	}
 	log.Println("connected to MongoDB")
 
+	// The unique indexes are a security property: GetUserByEmail does an
+	// unscoped FindOne, so duplicate (or blank) emails let the OIDC callback's
+	// cross-org guard compare against an arbitrary user, and duplicate org slugs
+	// make host-based org resolution pick one at random.
+	if err := services.EnsureAuthIndexes(); err != nil {
+		log.Fatalf("failed to ensure auth indexes: %v", err)
+	}
+	if err := services.RunMigrations(); err != nil {
+		log.Fatalf("migration failed: %v", err)
+	}
+	// Must run before the unique settings indexes are built, and before 0003:
+	// 0003 can create a "default" org, which would push 0002 into its ambiguous
+	// multi-org branch and leave the settings doc unstamped.
+	if err := services.MigrateSettingsOrg(); err != nil {
+		log.Fatalf("settings org migration failed: %v", err)
+	}
+	if err := services.MigrateMissedOrgScopes(); err != nil {
+		log.Fatalf("missed org scope migration failed: %v", err)
+	}
+
 	if err := services.EnsureSecretIndexes(); err != nil {
 		log.Printf("warning: failed to ensure secret indexes: %v", err)
 	}
 
+	// The unique indexes are a security property: duplicate settings docs make
+	// GetSettings return an arbitrary one, and duplicate ESO token hashes make
+	// ResolveSecretsReadToken pick an arbitrary org.
+	if err := services.EnsureSettingsIndexes(); err != nil {
+		log.Fatalf("failed to ensure settings indexes: %v", err)
+	}
+
 	if err := services.EnsureWorkflowIndexes(); err != nil {
 		log.Printf("warning: failed to ensure workflow indexes: %v", err)
 	}
 
+	if orgIDs, err := services.ListOrgIDs(); err != nil {
+		log.Printf("warning: failed to list orgs for default step seeding: %v", err)
+	} else {
+		for _, orgID := range orgIDs {
+			if created, updated, err := services.SeedDefaultSteps(orgID); err != nil {
+				log.Printf("warning: failed to seed default steps for org %s: %v", orgID, err)
+			} else {
+				log.Printf("default steps seeded for org %s: %d created, %d updated", orgID, created, updated)
+			}
+		}
+	}
+
 	services.StartLogSweeper()
 
 	redisAddr := getEnv("REDIS_ADDR", "localhost:6379")
@@ -39,10 +86,6 @@ func main() {
 	}
 	log.Println("connected to Redis")
 
-	if err := auth.InitOIDC(context.Background()); err != nil {
-		log.Fatalf("failed to initialise OIDC: %v", err)
-	}
-
 	// Background goroutine to mark offline servers
 	go func() {
 		ticker := time.NewTicker(2 * time.Minute)
@@ -61,6 +104,9 @@ func main() {
 		}
 	}()
 
+	// Start the server-side monitor scheduler.
+	monitorsched.Start(context.Background())
+
 	// Start REST server
 	r := gin.New()
 	r.Use(gin.Recovery())

server/internal/api/channels.go

@@ -0,0 +1,97 @@
+package api
+
+import (
+	"net/http"
+
+	"github.com/gin-gonic/gin"
+	"github.com/mrhid6/vantage/server/internal/auth"
+	"github.com/mrhid6/vantage/server/internal/models"
+	"github.com/mrhid6/vantage/server/internal/services"
+	"go.mongodb.org/mongo-driver/v2/bson"
+)
+
+func registerChannelRoutes(g *gin.RouterGroup) {
+	g.GET("/channels", listChannels)
+	g.POST("/channels", createChannel)
+	g.PUT("/channels/:id", updateChannel)
+	g.DELETE("/channels/:id", deleteChannel)
+	g.POST("/channels/:id/test", testChannel)
+}
+
+func listChannels(c *gin.Context) {
+	channels, err := services.ListChannels(auth.OrgID(c))
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusOK, channels)
+}
+
+func createChannel(c *gin.Context) {
+	var ch models.NotificationChannel
+	if err := c.ShouldBindJSON(&ch); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	if ch.Name == "" || ch.Type == "" {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "name and type are required"})
+		return
+	}
+	created, err := services.CreateChannel(auth.OrgID(c), &ch)
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusCreated, created)
+}
+
+func updateChannel(c *gin.Context) {
+	var body struct {
+		Name    *string            `json:"name"`
+		Type    *string            `json:"type"`
+		Config  *map[string]string `json:"config"`
+		Enabled *bool              `json:"enabled"`
+	}
+	if err := c.ShouldBindJSON(&body); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	upd := bson.M{}
+	if body.Name != nil {
+		upd["name"] = *body.Name
+	}
+	if body.Type != nil {
+		upd["type"] = *body.Type
+	}
+	if body.Config != nil {
+		upd["config"] = *body.Config
+	}
+	if body.Enabled != nil {
+		upd["enabled"] = *body.Enabled
+	}
+	if len(upd) == 0 {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "no fields to update"})
+		return
+	}
+	if err := services.UpdateChannel(auth.OrgID(c), c.Param("id"), upd); err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	c.Status(http.StatusNoContent)
+}
+
+func deleteChannel(c *gin.Context) {
+	if err := services.DeleteChannel(auth.OrgID(c), c.Param("id")); err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	c.Status(http.StatusNoContent)
+}
+
+func testChannel(c *gin.Context) {
+	if err := services.TestChannel(auth.OrgID(c), c.Param("id")); err != nil {
+		c.JSON(http.StatusBadGateway, gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusOK, gin.H{"status": "sent"})
+}

server/internal/api/console.go

@@ -8,6 +8,7 @@ import (
 	"time"
 
 	"github.com/gin-gonic/gin"
+	"github.com/mrhid6/vantage/server/internal/auth"
 	"github.com/mrhid6/vantage/server/internal/services"
 	"github.com/wwt/guac"
 )
@@ -29,13 +30,13 @@ func consoleConnect(c *gin.Context) {
 		return
 	}
 
-	srv, err := services.GetServer(body.ServerID)
+	srv, err := services.GetServer(auth.OrgID(c), body.ServerID)
 	if err != nil {
 		c.JSON(http.StatusNotFound, gin.H{"error": "server not found"})
 		return
 	}
 
-	sess, err := services.CreateConsoleSession(body.ServerID, body.Protocol, body.KeyID, actorFromCtx(c), c.ClientIP())
+	sess, err := services.CreateConsoleSession(auth.OrgID(c), body.ServerID, body.Protocol, body.KeyID, actorFromCtx(c), c.ClientIP())
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
@@ -47,20 +48,20 @@ func consoleConnect(c *gin.Context) {
 	}
 
 	if (body.Protocol == "rdp" || body.Protocol == "vnc") && (body.RDPUsername != "" || body.RDPPassword != "") {
-		if err := services.StashConsoleRDPCreds(sess.SessionID, body.RDPUsername, body.RDPPassword); err != nil {
+		if err := services.StashConsoleRDPCreds(auth.OrgID(c), sess.SessionID, body.RDPUsername, body.RDPPassword); err != nil {
 			c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 			return
 		}
 	}
 
 	if body.Protocol == "ssh" {
-		if err := services.SetConsoleSSHUser(sess.SessionID, body.SSHUsername); err != nil {
+		if err := services.SetConsoleSSHUser(auth.OrgID(c), sess.SessionID, body.SSHUsername); err != nil {
 			c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 			return
 		}
 	}
 
-	services.LogEvent("console.opened", actorFromCtx(c), srv.ServerID, "",
+	services.LogEvent(auth.OrgID(c), "console.opened", actorFromCtx(c), srv.ServerID, "",
 		"console session opened ("+body.Protocol+")")
 
 	c.JSON(http.StatusOK, gin.H{
@@ -88,7 +89,8 @@ func consoleTunnel(c *gin.Context) {
 		c.JSON(http.StatusUnauthorized, gin.H{"error": "invalid token"})
 		return
 	}
-	sess, err := services.GetConsoleSession(sessionID)
+	orgID := auth.OrgID(c)
+	sess, err := services.GetConsoleSession(orgID, sessionID)
 	if err != nil {
 		c.JSON(http.StatusNotFound, gin.H{"error": "session not found"})
 		return
@@ -102,12 +104,12 @@ func consoleTunnel(c *gin.Context) {
 	}
 
 	// Single-use: atomically spend the token so a replay within its TTL is rejected.
-	if err := services.ConsumeSessionToken(sessionID); err != nil {
+	if err := services.ConsumeSessionToken(orgID, sessionID); err != nil {
 		c.JSON(http.StatusUnauthorized, gin.H{"error": "token already used"})
 		return
 	}
 
-	srv, err := services.GetServer(sess.ServerID)
+	srv, err := services.GetServer(auth.OrgID(c), sess.ServerID)
 	if err != nil {
 		c.JSON(http.StatusNotFound, gin.H{"error": "server not found"})
 		return
@@ -116,7 +118,7 @@ func consoleTunnel(c *gin.Context) {
 	// Decrypt private key + passphrase in-memory only (ssh).
 	var privKey, passphrase string
 	if sess.Protocol == "ssh" && sess.KeyID != "" {
-		privKey, err = services.GetPrivateKey(sess.KeyID)
+		privKey, err = services.GetPrivateKey(auth.OrgID(c), sess.KeyID)
 		if err != nil {
 			c.JSON(http.StatusBadRequest, gin.H{"error": "selected key has no private material"})
 			return
@@ -126,7 +128,7 @@ func consoleTunnel(c *gin.Context) {
 
 	var rdpUser, rdpPass string
 	if sess.Protocol == "rdp" || sess.Protocol == "vnc" {
-		rdpUser, rdpPass, err = services.ConsumeConsoleRDPCreds(sessionID)
+		rdpUser, rdpPass, err = services.ConsumeConsoleRDPCreds(orgID, sessionID)
 		if err != nil {
 			c.JSON(http.StatusInternalServerError, gin.H{"error": "could not load credentials"})
 			return
@@ -171,7 +173,7 @@ func consoleTunnel(c *gin.Context) {
 
 	wsServer := guac.NewWebsocketServer(connect)
 	wsServer.OnDisconnect = func(id string, r *http.Request, t guac.Tunnel) {
-		_ = services.EndConsoleSession(sessionID)
+		_ = services.EndConsoleSession(orgID, sessionID)
 	}
 	wsServer.ServeHTTP(c.Writer, c.Request)
 }

server/internal/api/handlers.go

@@ -32,11 +32,14 @@ func RegisterRoutes(r *gin.Engine) {
 	// group as flat JSON.
 	r.GET("/api/secrets/:group/values", secretsReadAuth(), esoGetGroup)
 
-	// Auth endpoints (no session required)
-	r.GET("/auth/login", auth.HandleLogin)
-	r.GET("/auth/callback", auth.HandleCallback)
-	r.GET("/auth/logout", auth.HandleLogout)
+	// Unauthenticated auth endpoints
+	r.GET("/auth/bootstrap-status", auth.HandleBootstrapStatus)
+	r.POST("/auth/bootstrap", auth.HandleBootstrap)
+	r.POST("/auth/login", auth.HandleLocalLogin)
+	r.POST("/auth/logout", auth.HandleLogout)
 	r.GET("/auth/me", auth.HandleMe)
+	r.GET("/auth/oidc/start", auth.HandleOIDCStart)
+	r.GET("/auth/oidc/callback", auth.HandleOIDCCallback)
 
 	// API endpoints protected by session middleware
 	apiGroup := r.Group("/api")
@@ -56,9 +59,13 @@ func RegisterRoutes(r *gin.Engine) {
 
 		apiGroup.GET("/audit", listAuditEvents)
 
-		apiGroup.GET("/settings", getSettings)
-		apiGroup.PUT("/settings", saveSettings)
-		apiGroup.POST("/settings/secrets-token", rotateSecretsToken)
+		settings := apiGroup.Group("/settings")
+		settings.Use(auth.RequireRole("owner", "admin"))
+		{
+			settings.GET("", getSettings)
+			settings.PUT("", saveSettings)
+			settings.POST("/secrets-token", rotateSecretsToken)
+		}
 
 		apiGroup.GET("/secrets", listSecretGroups)
 		apiGroup.POST("/secrets", createSecretGroup)
@@ -80,11 +87,24 @@ func RegisterRoutes(r *gin.Engine) {
 		apiGroup.GET("/console/tunnel", consoleTunnel)
 
 		registerWorkflowRoutes(apiGroup)
+		registerMonitorRoutes(apiGroup)
+		registerChannelRoutes(apiGroup)
+
+		org := apiGroup.Group("/org")
+		org.Use(auth.RequireRole("owner", "admin"))
+		{
+			org.GET("/users", listOrgUsers)
+			org.POST("/users", createOrgUser)
+			org.PUT("/users/:id/role", updateOrgUserRole)
+			org.DELETE("/users/:id", deleteOrgUser)
+			org.GET("/oidc", getOrgOIDC)
+			org.PUT("/oidc", putOrgOIDC)
+		}
 	}
 }
 
 func listServers(c *gin.Context) {
-	servers, err := services.ListServers()
+	servers, err := services.ListServers(auth.OrgID(c))
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
@@ -93,7 +113,7 @@ func listServers(c *gin.Context) {
 }
 
 func createServer(c *gin.Context) {
-	s, token, err := services.CreateServer()
+	s, token, err := services.CreateServer(auth.OrgID(c))
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
@@ -106,21 +126,19 @@ func createServer(c *gin.Context) {
 }
 
 func newServer(c *gin.Context) {
-	s, token, err := services.CreateServer()
+	s, token, err := services.CreateServer(auth.OrgID(c))
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("server.created", actorFromCtx(c), s.ServerID, "", "pre-registration token issued")
+	services.LogEvent(auth.OrgID(c), "server.created", actorFromCtx(c), s.ServerID, "", "pre-registration token issued")
 
 	giteaHost := os.Getenv("GITEA_HOST")
 	if giteaHost == "" {
 		giteaHost = "gitea.example.com"
 	}
-	host := os.Getenv("PUBLIC_HOST")
-	if host == "" {
-		host = "https://vantage.example.com"
-	}
+
+	host := publicHostFromRequest(c)
 
 	installCmd := fmt.Sprintf(
 		`curl -fsSL "%s/install?server_id=%s&token=%s" | bash`,
@@ -142,13 +160,13 @@ func newServer(c *gin.Context) {
 
 func getServer(c *gin.Context) {
 	id := c.Param("id")
-	s, err := services.GetServer(id)
+	s, err := services.GetServer(auth.OrgID(c), id)
 	if err != nil {
 		c.JSON(http.StatusNotFound, gin.H{"error": "server not found"})
 		return
 	}
 
-	assignments, _ := services.GetAssignmentsWithKeysForServer(id)
+	assignments, _ := services.GetAssignmentsWithKeysForServer(auth.OrgID(c), id)
 
 	// Build response matching ServerWithKeys shape expected by frontend
 	type serverResponse struct {
@@ -163,8 +181,8 @@ func getServer(c *gin.Context) {
 
 func deleteServer(c *gin.Context) {
 	id := c.Param("id")
-	s, _ := services.GetServer(id)
-	if err := services.DeleteServer(id); err != nil {
+	s, _ := services.GetServer(auth.OrgID(c), id)
+	if err := services.DeleteServer(auth.OrgID(c), id); err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
@@ -172,7 +190,7 @@ func deleteServer(c *gin.Context) {
 	if s != nil {
 		hostname = s.Hostname
 	}
-	services.LogEvent("server.deleted", actorFromCtx(c), id, "", fmt.Sprintf("server %s deleted", hostname))
+	services.LogEvent(auth.OrgID(c), "server.deleted", actorFromCtx(c), id, "", fmt.Sprintf("server %s deleted", hostname))
 	c.JSON(http.StatusOK, gin.H{"deleted": true})
 }
 
@@ -191,7 +209,7 @@ func generateKey(c *gin.Context) {
 		body.Label = "generated"
 	}
 
-	s, err := services.GetServer(id)
+	s, err := services.GetServer(auth.OrgID(c), id)
 	if err != nil {
 		c.JSON(http.StatusNotFound, gin.H{"error": "server not found"})
 		return
@@ -209,7 +227,7 @@ func generateKey(c *gin.Context) {
 		return
 	}
 
-	services.LogEvent("key.generation_dispatched", actorFromCtx(c), s.ServerID, "", fmt.Sprintf("key generation dispatched (label=%s type=%s)", body.Label, body.KeyType))
+	services.LogEvent(auth.OrgID(c), "key.generation_dispatched", actorFromCtx(c), s.ServerID, "", fmt.Sprintf("key generation dispatched (label=%s type=%s)", body.Label, body.KeyType))
 	c.JSON(http.StatusAccepted, gin.H{
 		"message":    "key generation command sent to agent",
 		"command_id": cmdID,
@@ -218,7 +236,7 @@ func generateKey(c *gin.Context) {
 }
 
 func listKeys(c *gin.Context) {
-	keys, err := services.ListKeys()
+	keys, err := services.ListKeys(auth.OrgID(c))
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
@@ -238,18 +256,18 @@ func createKey(c *gin.Context) {
 		return
 	}
 
-	key, err := services.CreateKey(body.Label, body.PublicKey, "uploaded", "", body.PrivateKey, body.Passphrase)
+	key, err := services.CreateKey(auth.OrgID(c), body.Label, body.PublicKey, "uploaded", "", body.PrivateKey, body.Passphrase)
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("key.uploaded", actorFromCtx(c), "", key.KeyID, fmt.Sprintf("key '%s' uploaded", key.Label))
+	services.LogEvent(auth.OrgID(c), "key.uploaded", actorFromCtx(c), "", key.KeyID, fmt.Sprintf("key '%s' uploaded", key.Label))
 	c.JSON(http.StatusCreated, key)
 }
 
 func getPrivateKey(c *gin.Context) {
 	id := c.Param("id")
-	plaintext, err := services.GetPrivateKey(id)
+	plaintext, err := services.GetPrivateKey(auth.OrgID(c), id)
 	if err != nil {
 		c.JSON(http.StatusNotFound, gin.H{"error": err.Error()})
 		return
@@ -259,13 +277,13 @@ func getPrivateKey(c *gin.Context) {
 
 func getKey(c *gin.Context) {
 	id := c.Param("id")
-	key, err := services.GetKey(id)
+	key, err := services.GetKey(auth.OrgID(c), id)
 	if err != nil {
 		c.JSON(http.StatusNotFound, gin.H{"error": "key not found"})
 		return
 	}
 
-	assignments, _ := services.GetAssignmentsWithServers(id)
+	assignments, _ := services.GetAssignmentsWithServers(auth.OrgID(c), id)
 
 	type keyResponse struct {
 		*models.Key
@@ -279,8 +297,8 @@ func getKey(c *gin.Context) {
 
 func deleteKey(c *gin.Context) {
 	id := c.Param("id")
-	k, _ := services.GetKey(id)
-	if err := services.DeleteKey(id); err != nil {
+	k, _ := services.GetKey(auth.OrgID(c), id)
+	if err := services.DeleteKey(auth.OrgID(c), id); err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
@@ -288,7 +306,7 @@ func deleteKey(c *gin.Context) {
 	if k != nil {
 		label = k.Label
 	}
-	services.LogEvent("key.deleted", actorFromCtx(c), "", id, fmt.Sprintf("key '%s' deleted", label))
+	services.LogEvent(auth.OrgID(c), "key.deleted", actorFromCtx(c), "", id, fmt.Sprintf("key '%s' deleted", label))
 	c.JSON(http.StatusOK, gin.H{"deleted": true})
 }
 
@@ -302,12 +320,12 @@ func assignKey(c *gin.Context) {
 		return
 	}
 
-	a, err := services.AssignKey(keyID, body.ServerID)
+	a, err := services.AssignKey(auth.OrgID(c), keyID, body.ServerID)
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("key.assigned", actorFromCtx(c), body.ServerID, keyID, fmt.Sprintf("key %s assigned to server %s", keyID, body.ServerID))
+	services.LogEvent(auth.OrgID(c), "key.assigned", actorFromCtx(c), body.ServerID, keyID, fmt.Sprintf("key %s assigned to server %s", keyID, body.ServerID))
 	c.JSON(http.StatusCreated, a)
 }
 
@@ -315,11 +333,11 @@ func revokeAssignment(c *gin.Context) {
 	keyID := c.Param("id")
 	serverID := c.Param("serverId")
 
-	if err := services.RevokeAssignment(keyID, serverID); err != nil {
+	if err := services.RevokeAssignment(auth.OrgID(c), keyID, serverID); err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("key.revoked", actorFromCtx(c), serverID, keyID, fmt.Sprintf("key %s revoked from server %s", keyID, serverID))
+	services.LogEvent(auth.OrgID(c), "key.revoked", actorFromCtx(c), serverID, keyID, fmt.Sprintf("key %s revoked from server %s", keyID, serverID))
 	c.JSON(http.StatusOK, gin.H{"revoked": true})
 }
 
@@ -334,7 +352,7 @@ func getLatestAgentVersion(c *gin.Context) {
 
 func updateAgent(c *gin.Context) {
 	id := c.Param("id")
-	s, err := services.GetServer(id)
+	s, err := services.GetServer(auth.OrgID(c), id)
 	if err != nil {
 		c.JSON(http.StatusNotFound, gin.H{"error": "server not found"})
 		return
@@ -345,7 +363,7 @@ func updateAgent(c *gin.Context) {
 		c.JSON(http.StatusServiceUnavailable, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("agent.update_dispatched", actorFromCtx(c), s.ServerID, "", fmt.Sprintf("agent update dispatched to %s (version %s)", s.Hostname, version))
+	services.LogEvent(auth.OrgID(c), "agent.update_dispatched", actorFromCtx(c), s.ServerID, "", fmt.Sprintf("agent update dispatched to %s (version %s)", s.Hostname, version))
 	c.JSON(http.StatusAccepted, gin.H{
 		"message": "update command sent to agent",
 		"version": version,
@@ -354,7 +372,7 @@ func updateAgent(c *gin.Context) {
 
 func applyUpdates(c *gin.Context) {
 	id := c.Param("id")
-	s, err := services.GetServer(id)
+	s, err := services.GetServer(auth.OrgID(c), id)
 	if err != nil {
 		c.JSON(http.StatusNotFound, gin.H{"error": "server not found"})
 		return
@@ -364,7 +382,7 @@ func applyUpdates(c *gin.Context) {
 		c.JSON(http.StatusServiceUnavailable, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("updates.applied", actorFromCtx(c), s.ServerID, "", fmt.Sprintf("package update command dispatched to %s", s.Hostname))
+	services.LogEvent(auth.OrgID(c), "updates.applied", actorFromCtx(c), s.ServerID, "", fmt.Sprintf("package update command dispatched to %s", s.Hostname))
 	c.JSON(http.StatusAccepted, gin.H{"message": "apply updates command sent to agent"})
 }
 
@@ -431,7 +449,7 @@ func listAuditEvents(c *gin.Context) {
 			limit = n
 		}
 	}
-	events, err := services.ListAuditEvents(limit)
+	events, err := services.ListAuditEvents(auth.OrgID(c), limit)
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
@@ -440,7 +458,7 @@ func listAuditEvents(c *gin.Context) {
 }
 
 func getSettings(c *gin.Context) {
-	s, err := services.GetSettings()
+	s, err := services.GetSettings(auth.OrgID(c))
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
@@ -458,11 +476,11 @@ func saveSettings(c *gin.Context) {
 		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
 		return
 	}
-	if err := services.SaveSettings(body.Alerts, body.Email, body.WorkflowLogRetentionDays); err != nil {
+	if err := services.SaveSettings(auth.OrgID(c), body.Alerts, body.Email, body.WorkflowLogRetentionDays); err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("settings.updated", actorFromCtx(c), "", "", "alert settings updated")
+	services.LogEvent(auth.OrgID(c), "settings.updated", actorFromCtx(c), "", "", "alert settings updated")
 	c.JSON(http.StatusOK, gin.H{"saved": true})
 }
 
@@ -474,14 +492,7 @@ func handleInstallScript(c *gin.Context) {
 	if giteaHost == "" {
 		giteaHost = "gitea.example.com"
 	}
-	publicHost := os.Getenv("PUBLIC_HOST")
-	if publicHost == "" {
-		publicHost = "vantage.example.com"
-	}
 	grpcHost := os.Getenv("GRPC_HOST")
-	if grpcHost == "" {
-		grpcHost = publicHost
-	}
 
 	script := fmt.Sprintf(`#!/usr/bin/env bash
 set -euo pipefail
@@ -489,9 +500,6 @@ set -euo pipefail
 SERVER_ID="%s"
 TOKEN="%s"
 GITEA_HOST="%s"
-KM_HOST="%s"
-KM_HOST="${KM_HOST#https://}"
-KM_HOST="${KM_HOST#http://}"
 GRPC_HOST="%s"
 GRPC_HOST="${GRPC_HOST#https://}"
 GRPC_HOST="${GRPC_HOST#http://}"
@@ -564,7 +572,7 @@ systemctl daemon-reload
 systemctl enable --now vantage-agent
 
 echo "vantage-agent installed and started."
-`, serverID, token, giteaHost, publicHost, grpcHost)
+`, serverID, token, giteaHost, grpcHost)
 
 	c.Header("Content-Type", "text/x-shellscript")
 	c.String(http.StatusOK, script)

server/internal/api/install_ps1.go

@@ -16,13 +16,8 @@ func handleInstallScriptWindows(c *gin.Context) {
 	if giteaHost == "" {
 		giteaHost = "gitea.example.com"
 	}
+	// Guaranteed non-empty: main() fatals at boot if GRPC_HOST is unset.
 	grpcHost := os.Getenv("GRPC_HOST")
-	if grpcHost == "" {
-		grpcHost = os.Getenv("PUBLIC_HOST")
-	}
-	if grpcHost == "" {
-		grpcHost = "vantage.example.com"
-	}
 
 	script := fmt.Sprintf(
 		"#Requires -RunAsAdministrator\n"+

server/internal/api/monitors.go

@@ -0,0 +1,158 @@
+package api
+
+import (
+	"net/http"
+	"time"
+
+	"github.com/gin-gonic/gin"
+	"github.com/mrhid6/vantage/server/internal/auth"
+	"github.com/mrhid6/vantage/server/internal/models"
+	"github.com/mrhid6/vantage/server/internal/services"
+	"go.mongodb.org/mongo-driver/v2/bson"
+)
+
+func registerMonitorRoutes(g *gin.RouterGroup) {
+	g.GET("/monitors", listMonitors)
+	g.POST("/monitors", createMonitor)
+	g.GET("/monitors/:id", getMonitor)
+	g.PUT("/monitors/:id", updateMonitor)
+	g.DELETE("/monitors/:id", deleteMonitor)
+	g.GET("/monitors/:id/incidents", getMonitorIncidents)
+	g.GET("/monitors/:id/uptime", getMonitorUptime)
+}
+
+func listMonitors(c *gin.Context) {
+	monitors, err := services.ListMonitors(auth.OrgID(c))
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusOK, monitors)
+}
+
+func createMonitor(c *gin.Context) {
+	var m models.Monitor
+	if err := c.ShouldBindJSON(&m); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	if m.Name == "" || m.Type == "" {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "name and type are required"})
+		return
+	}
+	created, err := services.CreateMonitor(auth.OrgID(c), &m)
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusCreated, created)
+}
+
+func getMonitor(c *gin.Context) {
+	m, err := services.GetMonitor(auth.OrgID(c), c.Param("id"))
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	if m == nil {
+		c.JSON(http.StatusNotFound, gin.H{"error": "monitor not found"})
+		return
+	}
+	c.JSON(http.StatusOK, m)
+}
+
+func updateMonitor(c *gin.Context) {
+	var body struct {
+		Name        *string               `json:"name"`
+		Type        *string               `json:"type"`
+		Target      *models.MonitorTarget `json:"target"`
+		IntervalSec *int                  `json:"interval_sec"`
+		Runner      *string               `json:"runner"`
+		Retries     *int                  `json:"retries"`
+		Enabled     *bool                 `json:"enabled"`
+		ChannelIDs  *[]string             `json:"channel_ids"`
+	}
+	if err := c.ShouldBindJSON(&body); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	upd := bson.M{}
+	if body.Name != nil {
+		upd["name"] = *body.Name
+	}
+	if body.Type != nil {
+		upd["type"] = *body.Type
+	}
+	if body.Target != nil {
+		upd["target"] = *body.Target
+	}
+	if body.IntervalSec != nil {
+		upd["interval_sec"] = *body.IntervalSec
+	}
+	if body.Runner != nil {
+		upd["runner"] = *body.Runner
+	}
+	if body.Retries != nil {
+		upd["retries"] = *body.Retries
+	}
+	if body.Enabled != nil {
+		upd["enabled"] = *body.Enabled
+	}
+	if body.ChannelIDs != nil {
+		upd["channel_ids"] = *body.ChannelIDs
+	}
+	if len(upd) == 0 {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "no fields to update"})
+		return
+	}
+	if err := services.UpdateMonitor(auth.OrgID(c), c.Param("id"), upd); err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	c.Status(http.StatusNoContent)
+}
+
+func deleteMonitor(c *gin.Context) {
+	if err := services.DeleteMonitor(auth.OrgID(c), c.Param("id")); err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	c.Status(http.StatusNoContent)
+}
+
+func getMonitorIncidents(c *gin.Context) {
+	m, err := services.GetMonitor(auth.OrgID(c), c.Param("id"))
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	if m == nil {
+		c.JSON(http.StatusNotFound, gin.H{"error": "monitor not found"})
+		return
+	}
+	incidents, err := services.ListIncidents(auth.OrgID(c), c.Param("id"), 50)
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusOK, incidents)
+}
+
+func getMonitorUptime(c *gin.Context) {
+	m, err := services.GetMonitor(auth.OrgID(c), c.Param("id"))
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	if m == nil {
+		c.JSON(http.StatusNotFound, gin.H{"error": "monitor not found"})
+		return
+	}
+	since := time.Now().Add(-30 * 24 * time.Hour)
+	rollups, err := services.UptimeRollups(auth.OrgID(c), c.Param("id"), since)
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusOK, rollups)
+}

server/internal/api/org.go

@@ -0,0 +1,162 @@
+package api
+
+import (
+	"errors"
+	"net/http"
+
+	"github.com/gin-gonic/gin"
+	"github.com/mrhid6/vantage/server/internal/auth"
+	"github.com/mrhid6/vantage/server/internal/models"
+	"github.com/mrhid6/vantage/server/internal/services"
+)
+
+func listOrgUsers(c *gin.Context) {
+	users, err := services.ListUsers(auth.OrgID(c))
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusOK, users)
+}
+
+// Granting or removing the owner role is reserved to owners: an admin must
+// never be able to mint an owner (and log in as it) or strip the owners above
+// them. Everything below derives the actor from the session, never the body.
+func actorMayGrantOwner(c *gin.Context) bool {
+	return auth.Role(c) == models.RoleOwner
+}
+
+func createOrgUser(c *gin.Context) {
+	var body struct {
+		Email    string `json:"email"`
+		Password string `json:"password"`
+		Role     string `json:"role"`
+	}
+	if err := c.ShouldBindJSON(&body); err != nil || body.Email == "" {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "email required"})
+		return
+	}
+	if body.Role == "" {
+		body.Role = models.RoleMember
+	}
+	if !models.ValidRole(body.Role) {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "role must be one of owner, admin, or member"})
+		return
+	}
+	if body.Role == models.RoleOwner && !actorMayGrantOwner(c) {
+		c.JSON(http.StatusForbidden, gin.H{"error": "only an owner can create another owner"})
+		return
+	}
+	u, err := services.CreateUser(auth.OrgID(c), body.Email, body.Password, body.Role, "local")
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusCreated, u)
+}
+
+func updateOrgUserRole(c *gin.Context) {
+	var body struct {
+		Role string `json:"role"`
+	}
+	if err := c.ShouldBindJSON(&body); err != nil || body.Role == "" {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "role required"})
+		return
+	}
+	if !models.ValidRole(body.Role) {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "role must be one of owner, admin, or member"})
+		return
+	}
+
+	orgID, targetID := auth.OrgID(c), c.Param("id")
+	if targetID == auth.UserID(c) {
+		c.JSON(http.StatusForbidden, gin.H{"error": "you cannot change your own role"})
+		return
+	}
+	target, err := services.GetUserInOrg(orgID, targetID)
+	if err != nil {
+		c.JSON(http.StatusNotFound, gin.H{"error": "user not found"})
+		return
+	}
+	if (body.Role == models.RoleOwner || target.Role == models.RoleOwner) && !actorMayGrantOwner(c) {
+		c.JSON(http.StatusForbidden, gin.H{"error": "only an owner can change owner roles"})
+		return
+	}
+
+	if err := services.UpdateUserRole(orgID, targetID, body.Role); err != nil {
+		c.JSON(orgUserErrStatus(err), gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusOK, gin.H{"ok": true})
+}
+
+func deleteOrgUser(c *gin.Context) {
+	orgID, targetID := auth.OrgID(c), c.Param("id")
+	if targetID == auth.UserID(c) {
+		c.JSON(http.StatusForbidden, gin.H{"error": "you cannot remove your own account"})
+		return
+	}
+	target, err := services.GetUserInOrg(orgID, targetID)
+	if err != nil {
+		c.JSON(http.StatusNotFound, gin.H{"error": "user not found"})
+		return
+	}
+	if target.Role == models.RoleOwner && !actorMayGrantOwner(c) {
+		c.JSON(http.StatusForbidden, gin.H{"error": "only an owner can remove another owner"})
+		return
+	}
+
+	if err := services.DeleteUser(orgID, targetID); err != nil {
+		c.JSON(orgUserErrStatus(err), gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusOK, gin.H{"deleted": true})
+}
+
+// The last-owner guard is a rejected request, not a server fault — surface it
+// as 409 so the UI shows the message rather than a generic failure.
+func orgUserErrStatus(err error) int {
+	if errors.Is(err, services.ErrLastOwner) {
+		return http.StatusConflict
+	}
+	return http.StatusInternalServerError
+}
+
+func getOrgOIDC(c *gin.Context) {
+	cfg, err := services.GetOrgOIDC(auth.OrgID(c))
+	if err != nil {
+		c.JSON(http.StatusOK, gin.H{"enabled": false, "client_secret_set": false})
+		return
+	}
+	// The client secret itself is write-only (never serialized); expose only
+	// whether one is stored so the UI can say so without leaking it.
+	c.JSON(http.StatusOK, gin.H{
+		"org_id":            cfg.OrgID,
+		"issuer":            cfg.Issuer,
+		"client_id":         cfg.ClientID,
+		"enabled":           cfg.Enabled,
+		"updated_at":        cfg.UpdatedAt,
+		"client_secret_set": cfg.ClientSecretEnc != "",
+	})
+}
+
+func putOrgOIDC(c *gin.Context) {
+	var body struct {
+		Issuer       string `json:"issuer"`
+		ClientID     string `json:"client_id"`
+		ClientSecret string `json:"client_secret"`
+		Enabled      bool   `json:"enabled"`
+	}
+	if err := c.ShouldBindJSON(&body); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	if err := services.SaveOrgOIDC(auth.OrgID(c), body.Issuer, body.ClientID, body.ClientSecret, body.Enabled); err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	// Drop the cached provider so a rotated issuer takes effect immediately —
+	// an admin moving off a compromised IdP must not keep authenticating there.
+	auth.EvictOIDCProvider(auth.OrgID(c))
+	c.JSON(http.StatusOK, gin.H{"saved": true})
+}

server/internal/api/publichost.go

@@ -0,0 +1,51 @@
+package api
+
+import (
+	"net"
+	"strings"
+
+	"github.com/gin-gonic/gin"
+)
+
+func publicHostFromRequest(c *gin.Context) string {
+	host := c.Request.Host
+	if h := firstForwarded(c.GetHeader("X-Forwarded-Host")); h != "" {
+		host = h
+	}
+	if host == "" {
+		return "https://vantage.example.com"
+	}
+	return schemeFor(c, host) + "://" + host
+}
+
+func schemeFor(c *gin.Context, host string) string {
+	if p := firstForwarded(c.GetHeader("X-Forwarded-Proto")); p != "" {
+		return p
+	}
+	if c.Request.TLS != nil {
+		return "https"
+	}
+	if isLoopback(host) {
+		return "http"
+	}
+	return "https"
+}
+
+func firstForwarded(v string) string {
+	if v == "" {
+		return ""
+	}
+	return strings.TrimSpace(strings.Split(v, ",")[0])
+}
+
+func isLoopback(host string) bool {
+	h, _, err := net.SplitHostPort(host)
+	if err != nil {
+		h = host
+	}
+	if h == "localhost" || strings.HasSuffix(h, ".localhost") {
+		return true
+	}
+	ip := net.ParseIP(h)
+	return ip != nil && ip.IsLoopback()
+}

server/internal/api/secrets.go

@@ -7,6 +7,7 @@ import (
 	"strings"
 
 	"github.com/gin-gonic/gin"
+	"github.com/mrhid6/vantage/server/internal/auth"
 	"github.com/mrhid6/vantage/server/internal/services"
 )
 
@@ -18,19 +19,29 @@ func validName(s string) bool {
 	return s != "" && len(s) <= 128 && groupNamePattern.MatchString(s)
 }
 
-// secretsReadAuth validates the ESO bearer token on the public read endpoint.
+// ctxSecretsOrgKey carries the org resolved from the ESO bearer token.
+const ctxSecretsOrgKey = "km_secrets_org"
+
+// secretsReadAuth validates the ESO bearer token on the public read endpoint
+// and stashes the org the token belongs to.
+//
+// This is the one endpoint whose org does NOT come from the session or the
+// host: External Secrets Operator calls it machine-to-machine with no session,
+// so the token itself is the org-bearing credential.
 func secretsReadAuth() gin.HandlerFunc {
 	return func(c *gin.Context) {
 		const prefix = "Bearer "
-		auth := c.GetHeader("Authorization")
-		if len(auth) <= len(prefix) || !strings.EqualFold(auth[:len(prefix)], prefix) {
+		authHeader := c.GetHeader("Authorization")
+		if len(authHeader) <= len(prefix) || !strings.EqualFold(authHeader[:len(prefix)], prefix) {
 			c.AbortWithStatusJSON(http.StatusUnauthorized, gin.H{"error": "missing bearer token"})
 			return
 		}
-		if !services.VerifySecretsReadToken(auth[len(prefix):]) {
+		orgID, ok := services.ResolveSecretsReadToken(authHeader[len(prefix):])
+		if !ok {
 			c.AbortWithStatusJSON(http.StatusUnauthorized, gin.H{"error": "invalid token"})
 			return
 		}
+		c.Set(ctxSecretsOrgKey, orgID)
 		c.Next()
 	}
 }
@@ -40,7 +51,14 @@ func secretsReadAuth() gin.HandlerFunc {
 // (ESO treats 404 as "deleted").
 func esoGetGroup(c *gin.Context) {
 	group := c.Param("group")
-	values, err := services.GetSecretGroupDecrypted(group)
+	// Org comes from the bearer token (set by secretsReadAuth), not a session.
+	orgID := c.GetString(ctxSecretsOrgKey)
+	if orgID == "" {
+		// Defence in depth: never query the store unscoped.
+		c.AbortWithStatusJSON(http.StatusUnauthorized, gin.H{"error": "invalid token"})
+		return
+	}
+	values, err := services.GetSecretGroupDecrypted(orgID, group)
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": "store error"})
 		return
@@ -53,7 +71,7 @@ func esoGetGroup(c *gin.Context) {
 }
 
 func listSecretGroups(c *gin.Context) {
-	groups, err := services.ListSecretGroups()
+	groups, err := services.ListSecretGroups(auth.OrgID(c))
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
@@ -86,17 +104,17 @@ func createSecretGroup(c *gin.Context) {
 			return
 		}
 	}
-	if err := services.UpsertSecrets(body.Group, body.Values); err != nil {
+	if err := services.UpsertSecrets(auth.OrgID(c), body.Group, body.Values); err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("secret.updated", actorFromCtx(c), "", "", fmt.Sprintf("group '%s' created with keys: %s", body.Group, strings.Join(services.SortedKeys(body.Values), ", ")))
+	services.LogEvent(auth.OrgID(c), "secret.updated", actorFromCtx(c), "", "", fmt.Sprintf("group '%s' created with keys: %s", body.Group, strings.Join(services.SortedKeys(body.Values), ", ")))
 	c.JSON(http.StatusCreated, gin.H{"group": body.Group})
 }
 
 func getSecretGroup(c *gin.Context) {
 	group := c.Param("group")
-	secrets, err := services.GetSecretGroup(group)
+	secrets, err := services.GetSecretGroup(auth.OrgID(c), group)
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
@@ -130,11 +148,11 @@ func putSecretGroup(c *gin.Context) {
 			return
 		}
 	}
-	if err := services.UpsertSecrets(group, values); err != nil {
+	if err := services.UpsertSecrets(auth.OrgID(c), group, values); err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("secret.updated", actorFromCtx(c), "", "", fmt.Sprintf("group '%s' keys updated: %s", group, strings.Join(services.SortedKeys(values), ", ")))
+	services.LogEvent(auth.OrgID(c), "secret.updated", actorFromCtx(c), "", "", fmt.Sprintf("group '%s' keys updated: %s", group, strings.Join(services.SortedKeys(values), ", ")))
 	c.JSON(http.StatusOK, gin.H{"saved": true})
 }
 
@@ -147,42 +165,42 @@ func revealSecret(c *gin.Context) {
 		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
 		return
 	}
-	value, err := services.RevealSecret(group, body.Key)
+	value, err := services.RevealSecret(auth.OrgID(c), group, body.Key)
 	if err != nil {
 		c.JSON(http.StatusNotFound, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("secret.revealed", actorFromCtx(c), "", "", fmt.Sprintf("value of '%s/%s' revealed", group, body.Key))
+	services.LogEvent(auth.OrgID(c), "secret.revealed", actorFromCtx(c), "", "", fmt.Sprintf("value of '%s/%s' revealed", group, body.Key))
 	c.JSON(http.StatusOK, gin.H{"value": value})
 }
 
 func deleteSecretKey(c *gin.Context) {
 	group := c.Param("group")
 	key := c.Param("key")
-	if err := services.DeleteSecret(group, key); err != nil {
+	if err := services.DeleteSecret(auth.OrgID(c), group, key); err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("secret.deleted", actorFromCtx(c), "", "", fmt.Sprintf("key '%s' deleted from group '%s'", key, group))
+	services.LogEvent(auth.OrgID(c), "secret.deleted", actorFromCtx(c), "", "", fmt.Sprintf("key '%s' deleted from group '%s'", key, group))
 	c.JSON(http.StatusOK, gin.H{"deleted": true})
 }
 
 func deleteSecretGroup(c *gin.Context) {
 	group := c.Param("group")
-	if err := services.DeleteSecretGroup(group); err != nil {
+	if err := services.DeleteSecretGroup(auth.OrgID(c), group); err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("secretgroup.deleted", actorFromCtx(c), "", "", fmt.Sprintf("group '%s' deleted", group))
+	services.LogEvent(auth.OrgID(c), "secretgroup.deleted", actorFromCtx(c), "", "", fmt.Sprintf("group '%s' deleted", group))
 	c.JSON(http.StatusOK, gin.H{"deleted": true})
 }
 
 func rotateSecretsToken(c *gin.Context) {
-	token, err := services.RotateSecretsReadToken()
+	token, err := services.RotateSecretsReadToken(auth.OrgID(c))
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("secrets.token_rotated", actorFromCtx(c), "", "", "ESO read token rotated")
+	services.LogEvent(auth.OrgID(c), "secrets.token_rotated", actorFromCtx(c), "", "", "ESO read token rotated")
 	c.JSON(http.StatusOK, gin.H{"token": token})
 }

server/internal/api/workflows.go

@@ -2,6 +2,7 @@ package api
 
 import (
 	"fmt"
+	"io"
 	"net/http"
 	"os"
 	"regexp"
@@ -10,6 +11,7 @@ import (
 	"time"
 
 	"github.com/gin-gonic/gin"
+	"github.com/mrhid6/vantage/server/internal/auth"
 	"github.com/mrhid6/vantage/server/internal/models"
 	"github.com/mrhid6/vantage/server/internal/services"
 )
@@ -19,6 +21,11 @@ func registerWorkflowRoutes(g *gin.RouterGroup) {
 	g.POST("/steps", createStep)
 	g.PUT("/steps/:id", updateStep)
 	g.DELETE("/steps/:id", deleteStep)
+	g.GET("/steps/:id/export", exportStep)
+	g.POST("/steps/import", importStep)
+	g.POST("/steps/seed-defaults", seedDefaults)
+	g.GET("/steps/usage", stepUsage)
+	g.POST("/steps/parse", parseStep)
 
 	g.GET("/workflows", listWorkflows)
 	g.POST("/workflows", createWorkflow)
@@ -99,9 +106,10 @@ func streamServerRunLog(c *gin.Context) {
 	ctx := c.Request.Context()
 	ticker := time.NewTicker(500 * time.Millisecond)
 	defer ticker.Stop()
+	orgID := auth.OrgID(c)
 	for {
 		sendNew()
-		if serverRunTerminal(runID, serverID) {
+		if serverRunTerminal(orgID, runID, serverID) {
 			sendNew() // final drain
 			_, _ = c.Writer.WriteString("event: done\ndata: end\n\n")
 			flusher.Flush()
@@ -116,8 +124,8 @@ func streamServerRunLog(c *gin.Context) {
 }
 
 // serverRunTerminal reports whether the given server-run has reached a terminal status.
-func serverRunTerminal(runID, serverID string) bool {
-	r, err := services.GetRun(runID)
+func serverRunTerminal(orgID, runID, serverID string) bool {
+	r, err := services.GetRun(orgID, runID)
 	if err != nil {
 		return true
 	}
@@ -141,7 +149,7 @@ func splitSSE(b []byte) []string {
 }
 
 func listSteps(c *gin.Context) {
-	steps, err := services.ListSteps()
+	steps, err := services.ListSteps(auth.OrgID(c))
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
@@ -149,18 +157,27 @@ func listSteps(c *gin.Context) {
 	c.JSON(http.StatusOK, steps)
 }
 
+func stepUsage(c *gin.Context) {
+	counts, err := services.StepUsageCounts(auth.OrgID(c))
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusOK, counts)
+}
+
 func createStep(c *gin.Context) {
 	var s models.WorkflowStep
 	if err := c.ShouldBindJSON(&s); err != nil {
 		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
 		return
 	}
-	out, err := services.CreateStep(s)
+	out, err := services.CreateStep(auth.OrgID(c), s)
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("workflow.step_created", actorFromCtx(c), "", out.StepID, fmt.Sprintf("step '%s' created", out.Name))
+	services.LogEvent(auth.OrgID(c), "workflow.step_created", actorFromCtx(c), "", out.StepID, fmt.Sprintf("step '%s' created", out.Name))
 	c.JSON(http.StatusCreated, out)
 }
 
@@ -170,25 +187,80 @@ func updateStep(c *gin.Context) {
 		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
 		return
 	}
-	if err := services.UpdateStep(c.Param("id"), s); err != nil {
+	if err := services.UpdateStep(auth.OrgID(c), c.Param("id"), s); err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("workflow.step_updated", actorFromCtx(c), "", c.Param("id"), "step updated")
+	services.LogEvent(auth.OrgID(c), "workflow.step_updated", actorFromCtx(c), "", c.Param("id"), "step updated")
 	c.JSON(http.StatusOK, gin.H{"updated": true})
 }
 
 func deleteStep(c *gin.Context) {
-	if err := services.DeleteStep(c.Param("id")); err != nil {
+	if err := services.DeleteStep(auth.OrgID(c), c.Param("id")); err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("workflow.step_deleted", actorFromCtx(c), "", c.Param("id"), "step deleted")
+	services.LogEvent(auth.OrgID(c), "workflow.step_deleted", actorFromCtx(c), "", c.Param("id"), "step deleted")
 	c.JSON(http.StatusOK, gin.H{"deleted": true})
 }
 
+func exportStep(c *gin.Context) {
+	b, err := services.ExportStep(auth.OrgID(c), c.Param("id"))
+	if err != nil {
+		c.JSON(http.StatusNotFound, gin.H{"error": err.Error()})
+		return
+	}
+	c.Header("Content-Disposition", fmt.Sprintf("attachment; filename=step-%s.json", c.Param("id")))
+	c.Data(http.StatusOK, "application/json", b)
+}
+
+func seedDefaults(c *gin.Context) {
+	created, updated, err := services.SeedDefaultSteps(auth.OrgID(c))
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	services.LogEvent(auth.OrgID(c), "workflow.defaults_synced", actorFromCtx(c), "", "", fmt.Sprintf("default steps synced: %d created, %d updated", created, updated))
+	c.JSON(http.StatusOK, gin.H{"created": created, "updated": updated})
+}
+
+const maxStepBodyBytes = 1 << 20 // 1 MiB
+
+func importStep(c *gin.Context) {
+	c.Request.Body = http.MaxBytesReader(c.Writer, c.Request.Body, maxStepBodyBytes)
+	body, err := io.ReadAll(c.Request.Body)
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	out, err := services.ImportStepToLibrary(auth.OrgID(c), body)
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	services.LogEvent(auth.OrgID(c), "workflow.step_imported", actorFromCtx(c), "", out.StepID, fmt.Sprintf("step '%s' imported", out.Name))
+	c.JSON(http.StatusCreated, out)
+}
+
+// parseStep validates a step doc and returns the normalized step WITHOUT
+// persisting — used by the editor to insert an imported ad-hoc (inline) step.
+func parseStep(c *gin.Context) {
+	c.Request.Body = http.MaxBytesReader(c.Writer, c.Request.Body, maxStepBodyBytes)
+	body, err := io.ReadAll(c.Request.Body)
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	s, err := services.ParseStepDoc(body)
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusOK, s)
+}
+
 func listWorkflows(c *gin.Context) {
-	wfs, err := services.ListWorkflows()
+	wfs, err := services.ListWorkflows(auth.OrgID(c))
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
@@ -202,17 +274,17 @@ func createWorkflow(c *gin.Context) {
 		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
 		return
 	}
-	out, err := services.CreateWorkflow(w)
+	out, err := services.CreateWorkflow(auth.OrgID(c), w)
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("workflow.created", actorFromCtx(c), "", out.WorkflowID, fmt.Sprintf("workflow '%s' created", out.Name))
+	services.LogEvent(auth.OrgID(c), "workflow.created", actorFromCtx(c), "", out.WorkflowID, fmt.Sprintf("workflow '%s' created", out.Name))
 	c.JSON(http.StatusCreated, out)
 }
 
 func getWorkflow(c *gin.Context) {
-	w, err := services.GetWorkflow(c.Param("id"))
+	w, err := services.GetWorkflow(auth.OrgID(c), c.Param("id"))
 	if err != nil {
 		c.JSON(http.StatusNotFound, gin.H{"error": err.Error()})
 		return
@@ -226,12 +298,12 @@ func updateWorkflow(c *gin.Context) {
 		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
 		return
 	}
-	if err := services.UpdateWorkflow(c.Param("id"), w); err != nil {
+	if err := services.UpdateWorkflow(auth.OrgID(c), c.Param("id"), w); err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("workflow.updated", actorFromCtx(c), "", c.Param("id"), "workflow updated")
-	updated, err := services.GetWorkflow(c.Param("id"))
+	services.LogEvent(auth.OrgID(c), "workflow.updated", actorFromCtx(c), "", c.Param("id"), "workflow updated")
+	updated, err := services.GetWorkflow(auth.OrgID(c), c.Param("id"))
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
@@ -240,21 +312,21 @@ func updateWorkflow(c *gin.Context) {
 }
 
 func deleteWorkflow(c *gin.Context) {
-	if err := services.DeleteWorkflow(c.Param("id")); err != nil {
+	if err := services.DeleteWorkflow(auth.OrgID(c), c.Param("id")); err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("workflow.deleted", actorFromCtx(c), "", c.Param("id"), "workflow deleted")
+	services.LogEvent(auth.OrgID(c), "workflow.deleted", actorFromCtx(c), "", c.Param("id"), "workflow deleted")
 	c.JSON(http.StatusOK, gin.H{"deleted": true})
 }
 
 func runWorkflow(c *gin.Context) {
-	runID, err := services.TriggerWorkflow(c.Param("id"), actorFromCtx(c))
+	runID, err := services.TriggerWorkflow(auth.OrgID(c), c.Param("id"), actorFromCtx(c))
 	if err != nil {
 		c.JSON(http.StatusServiceUnavailable, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("workflow.run_triggered", actorFromCtx(c), "", c.Param("id"), fmt.Sprintf("run %s triggered", runID))
+	services.LogEvent(auth.OrgID(c), "workflow.run_triggered", actorFromCtx(c), "", c.Param("id"), fmt.Sprintf("run %s triggered", runID))
 	c.JSON(http.StatusAccepted, gin.H{"run_id": runID})
 }
 
@@ -265,7 +337,7 @@ func listWorkflowRuns(c *gin.Context) {
 			limit = n
 		}
 	}
-	runs, err := services.ListRuns(c.Param("id"), limit)
+	runs, err := services.ListRuns(auth.OrgID(c), c.Param("id"), limit)
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
@@ -274,7 +346,7 @@ func listWorkflowRuns(c *gin.Context) {
 }
 
 func getRun(c *gin.Context) {
-	r, err := services.GetRun(c.Param("runId"))
+	r, err := services.GetRun(auth.OrgID(c), c.Param("runId"))
 	if err != nil {
 		c.JSON(http.StatusNotFound, gin.H{"error": err.Error()})
 		return
@@ -283,10 +355,10 @@ func getRun(c *gin.Context) {
 }
 
 func cancelRun(c *gin.Context) {
-	if err := services.CancelRun(c.Param("runId")); err != nil {
+	if err := services.CancelRun(auth.OrgID(c), c.Param("runId")); err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	services.LogEvent("workflow.run_cancelled", actorFromCtx(c), "", c.Param("runId"), "run cancelled")
+	services.LogEvent(auth.OrgID(c), "workflow.run_cancelled", actorFromCtx(c), "", c.Param("runId"), "run cancelled")
 	c.JSON(http.StatusOK, gin.H{"cancelled": true})
 }

server/internal/auth/local.go

@@ -0,0 +1,165 @@
+package auth
+
+import (
+	"fmt"
+	"net/http"
+
+	"github.com/gin-gonic/gin"
+	"github.com/mrhid6/vantage/server/internal/models"
+	"github.com/mrhid6/vantage/server/internal/services"
+)
+
+func SetSessionCookie(c *gin.Context, sessionID string) {
+	secure := c.Request.TLS != nil || c.GetHeader("X-Forwarded-Proto") == "https"
+	http.SetCookie(c.Writer, &http.Cookie{
+		Name:     sessionCookieName,
+		Value:    sessionID,
+		Path:     "/",
+		HttpOnly: true,
+		Secure:   secure,
+		SameSite: http.SameSiteLaxMode,
+		MaxAge:   int(sessionTTL.Seconds()),
+	})
+}
+
+func HandleLocalLogin(c *gin.Context) {
+	var body struct {
+		Email    string `json:"email"`
+		Password string `json:"password"`
+	}
+	if err := c.ShouldBindJSON(&body); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "email and password required"})
+		return
+	}
+	u, err := services.GetUserByEmail(body.Email)
+	if err != nil || !services.VerifyPassword(u, body.Password) {
+		c.JSON(http.StatusUnauthorized, gin.H{"error": "invalid credentials"})
+		return
+	}
+	sessionID, err := SaveSession(c.Request.Context(), &Session{
+		UserID: u.UserID, OrgID: u.OrgID, Role: u.Role, Email: u.Email,
+	})
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "session save failed"})
+		return
+	}
+	_ = services.TouchLastLogin(u.UserID)
+	SetSessionCookie(c, sessionID)
+	c.JSON(http.StatusOK, gin.H{"ok": true})
+}
+
+// HandleBootstrapStatus answers "does the caller need to run setup". It is
+// unauthenticated, so it must not report instance-wide state to whoever asks:
+// on an org host the answer is scoped to that org, and only the apex — the
+// genuine first-run entry point — gets the global "no users anywhere" answer.
+func HandleBootstrapStatus(c *gin.Context) {
+	var (
+		n   int64
+		err error
+	)
+	if org, ok := OrgFromHost(c); ok {
+		n, err = services.CountOrgUsers(org.OrgID)
+	} else {
+		n, err = services.CountUsers()
+	}
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusOK, gin.H{"needs_setup": n == 0})
+}
+
+// HandleBootstrap creates the very first org and its owner, so its guard stays
+// deliberately global: it may run once on an empty instance and never again,
+// regardless of which host it is called on.
+func HandleBootstrap(c *gin.Context) {
+	n, err := services.CountUsers()
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	if n > 0 {
+		c.JSON(http.StatusConflict, gin.H{"error": "setup already complete"})
+		return
+	}
+	var body struct {
+		OrgName  string `json:"org_name"`
+		Email    string `json:"email"`
+		Password string `json:"password"`
+	}
+	if err := c.ShouldBindJSON(&body); err != nil || body.OrgName == "" || body.Email == "" || len(body.Password) < 8 {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "org_name, email, and password (>=8 chars) required"})
+		return
+	}
+	// An upgrade from single-tenant arrives here with no users but with the org
+	// the migrations created and stamped onto every legacy document. Creating a
+	// second org would put the owner somewhere else entirely, and since every
+	// org-scoped read filters on org_id, the operator would land in an empty
+	// Vantage with all their real data still under the migrated org — silent,
+	// total-looking data loss. So adopt the existing org instead, and only
+	// create when there genuinely is none. Same `switch orgCount` shape as
+	// migration 0002.
+	orgCount, err := services.CountOrgs()
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	var org *models.Org
+	switch orgCount {
+	case 0:
+		org, err = services.CreateOrg(body.OrgName)
+	case 1:
+		var existing *models.Org
+		existing, err = services.FirstOrg()
+		if err == nil {
+			org, err = services.AdoptOrg(existing.OrgID, body.OrgName)
+		}
+	default:
+		c.JSON(http.StatusConflict, gin.H{"error": fmt.Sprintf(
+			"cannot bootstrap: %d organizations already exist but no users do; "+
+				"create the owner against the intended org rather than through setup, "+
+				"or remove the unintended orgs and retry", orgCount)})
+		return
+	}
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	u, err := services.CreateUser(org.OrgID, body.Email, body.Password, "owner", "local")
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	sessionID, err := SaveSession(c.Request.Context(), &Session{
+		UserID: u.UserID, OrgID: u.OrgID, Role: u.Role, Email: u.Email,
+	})
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "session save failed"})
+		return
+	}
+	SetSessionCookie(c, sessionID)
+	c.JSON(http.StatusCreated, gin.H{"org": org, "slug": org.Slug})
+}
+
+func HandleMe(c *gin.Context) {
+	cookie, err := c.Request.Cookie(sessionCookieName)
+	if err != nil {
+		c.JSON(http.StatusUnauthorized, gin.H{"error": "not authenticated"})
+		return
+	}
+	sess, err := GetSession(c.Request.Context(), cookie.Value)
+	if err != nil {
+		c.JSON(http.StatusUnauthorized, gin.H{"error": "session expired"})
+		return
+	}
+	// /auth/me is registered outside Middleware, so it repeats the middleware's
+	// host/org match itself. Without this, a session for org A presented on org
+	// B's host would render the shell while every /api call 403s.
+	if hostOrg, ok := OrgFromHost(c); ok && hostOrg.OrgID != sess.OrgID {
+		c.JSON(http.StatusForbidden, gin.H{"error": "org host mismatch"})
+		return
+	}
+
+	org, _ := services.GetOrg(sess.OrgID)
+	c.JSON(http.StatusOK, gin.H{"user": sess, "org": org})
+}

server/internal/auth/middleware.go

@@ -14,13 +14,42 @@ func GetSessionFromContext(c *gin.Context) *Session {
 	return sess
 }
 
+func OrgID(c *gin.Context) string {
+	if s := GetSessionFromContext(c); s != nil {
+		return s.OrgID
+	}
+	return ""
+}
+
+func Role(c *gin.Context) string {
+	if s := GetSessionFromContext(c); s != nil {
+		return s.Role
+	}
+	return ""
+}
+
+func UserID(c *gin.Context) string {
+	if s := GetSessionFromContext(c); s != nil {
+		return s.UserID
+	}
+	return ""
+}
+
+func RequireRole(roles ...string) gin.HandlerFunc {
+	return func(c *gin.Context) {
+		r := Role(c)
+		for _, want := range roles {
+			if r == want {
+				c.Next()
+				return
+			}
+		}
+		c.AbortWithStatusJSON(http.StatusForbidden, gin.H{"error": "insufficient role"})
+	}
+}
+
 func Middleware() gin.HandlerFunc {
 	return func(c *gin.Context) {
-		if !authEnabled {
-			c.Next()
-			return
-		}
-
 		cookie, err := c.Request.Cookie(sessionCookieName)
 		if err != nil {
 			c.AbortWithStatusJSON(http.StatusUnauthorized, gin.H{"error": "not authenticated"})
@@ -33,7 +62,20 @@ func Middleware() gin.HandlerFunc {
 			return
 		}
 
+			// An org-less session would turn every downstream scope into
+		// {"org_id": ""} — fail closed rather than query across tenants.
+		if sess.OrgID == "" {
+			c.AbortWithStatusJSON(http.StatusUnauthorized, gin.H{"error": "session has no organization"})
+			return
+		}
+
 		c.Set(ctxSessionKey, sess)
+
+		if hostOrg, ok := OrgFromHost(c); ok && hostOrg.OrgID != sess.OrgID {
+			c.AbortWithStatusJSON(http.StatusForbidden, gin.H{"error": "org host mismatch"})
+			return
+		}
+
 		c.Next()
 	}
 }

server/internal/auth/oidc.go

@@ -2,123 +2,155 @@ package auth
 
 import (
 	"context"
-	"log"
+	"fmt"
 	"net/http"
-	"os"
+	"strings"
+	"sync"
 
 	"github.com/coreos/go-oidc/v3/oidc"
 	"github.com/gin-gonic/gin"
+	"github.com/mrhid6/vantage/server/internal/services"
 	"golang.org/x/oauth2"
 )
 
 var (
-	oidcProvider *oidc.Provider
-	oauth2Cfg    *oauth2.Config
-	authEnabled  bool
+	provMu    sync.Mutex
+	provCache = map[string]*oidc.Provider{}
 )
 
-func InitOIDC(ctx context.Context) error {
-	issuer := os.Getenv("OIDC_ISSUER")
-	if issuer == "" {
-		log.Println("OIDC_ISSUER not set; authentication disabled")
-		return nil
-	}
-
-	p, err := oidc.NewProvider(ctx, issuer)
-	if err != nil {
-		return err
-	}
-	oidcProvider = p
-	oauth2Cfg = &oauth2.Config{
-		ClientID:     os.Getenv("OIDC_CLIENT_ID"),
-		ClientSecret: os.Getenv("OIDC_CLIENT_SECRET"),
-		RedirectURL:  os.Getenv("OIDC_REDIRECT_URL"),
-		Endpoint:     p.Endpoint(),
-		Scopes:       []string{oidc.ScopeOpenID, "profile", "email"},
-	}
-	authEnabled = true
-	log.Println("OIDC authentication enabled")
-	return nil
+// EvictOIDCProvider drops an org's cached provider so the next login rediscovers
+// it from the (possibly changed) issuer. Called by the API layer after the org's
+// OIDC config is saved — services cannot import auth, so the handler wires it.
+func EvictOIDCProvider(orgID string) {
+	provMu.Lock()
+	delete(provCache, orgID)
+	provMu.Unlock()
 }
 
-func Enabled() bool { return authEnabled }
+func redirectURL(c *gin.Context) string {
+	scheme := "https"
+	if c.Request.TLS == nil && c.GetHeader("X-Forwarded-Proto") != "https" {
+		scheme = "http"
+	}
+	return fmt.Sprintf("%s://%s/auth/oidc/callback", scheme, c.Request.Host)
+}
 
-func HandleLogin(c *gin.Context) {
-	state, err := randomHex(16)
+// providerForOrg returns the org's (cached) OIDC provider plus a request-local
+// oauth2 config. The config is never stored on the cached entry: RedirectURL is
+// derived from this request's Host, so sharing it would let one in-flight login
+// overwrite another's redirect URI.
+func providerForOrg(ctx context.Context, c *gin.Context, orgID string) (*oidc.Provider, *oauth2.Config, error) {
+	cfg, err := services.GetOrgOIDC(orgID)
+	if err != nil || !cfg.Enabled {
+		return nil, nil, fmt.Errorf("org SSO not configured")
+	}
+	secret, err := services.GetOrgOIDCSecret(orgID)
 	if err != nil {
-		c.JSON(http.StatusInternalServerError, gin.H{"error": "state generation failed"})
+		return nil, nil, err
+	}
+	provMu.Lock()
+	p := provCache[orgID]
+	provMu.Unlock()
+	if p == nil {
+		p, err = oidc.NewProvider(ctx, cfg.Issuer)
+		if err != nil {
+			return nil, nil, err
+		}
+		provMu.Lock()
+		provCache[orgID] = p
+		provMu.Unlock()
+	}
+	return p, &oauth2.Config{
+		ClientID: cfg.ClientID, ClientSecret: secret,
+		RedirectURL: redirectURL(c), Endpoint: p.Endpoint(),
+		Scopes: []string{oidc.ScopeOpenID, "profile", "email"},
+	}, nil
+}
+
+func HandleOIDCStart(c *gin.Context) {
+	org, ok := OrgFromHost(c)
+	if !ok {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "unknown organization host"})
 		return
 	}
-	if err := SaveState(c.Request.Context(), state); err != nil {
+	ctx := c.Request.Context()
+	_, oauthCfg, err := providerForOrg(ctx, c, org.OrgID)
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	state, err := randomHex(16)
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "state gen failed"})
+		return
+	}
+	if err := SaveStateOrg(ctx, state, org.OrgID); err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": "state save failed"})
 		return
 	}
-	c.Redirect(http.StatusFound, oauth2Cfg.AuthCodeURL(state))
+	c.Redirect(http.StatusFound, oauthCfg.AuthCodeURL(state))
 }
 
-func HandleCallback(c *gin.Context) {
+func HandleOIDCCallback(c *gin.Context) {
 	ctx := c.Request.Context()
-
-	if !ConsumeState(ctx, c.Query("state")) {
+	orgID, ok := ConsumeStateOrg(ctx, c.Query("state"))
+	if !ok {
 		c.JSON(http.StatusBadRequest, gin.H{"error": "invalid state"})
 		return
 	}
-
-	token, err := oauth2Cfg.Exchange(ctx, c.Query("code"))
+	provider, oauthCfg, err := providerForOrg(ctx, c, orgID)
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	token, err := oauthCfg.Exchange(ctx, c.Query("code"))
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": "token exchange failed"})
 		return
 	}
-
 	rawIDToken, ok := token.Extra("id_token").(string)
 	if !ok {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": "missing id_token"})
 		return
 	}
-
-	verifier := oidcProvider.Verifier(&oidc.Config{ClientID: oauth2Cfg.ClientID})
-	idToken, err := verifier.Verify(ctx, rawIDToken)
+	idToken, err := provider.Verifier(&oidc.Config{ClientID: oauthCfg.ClientID}).Verify(ctx, rawIDToken)
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": "token verification failed"})
 		return
 	}
-
 	var claims struct {
-		Sub   string `json:"sub"`
 		Email string `json:"email"`
 		Name  string `json:"name"`
 	}
-	if err := idToken.Claims(&claims); err != nil {
+	if err := idToken.Claims(&claims); err != nil || claims.Email == "" {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": "claims extraction failed"})
 		return
 	}
 
+	email := strings.ToLower(claims.Email)
+	u, err := services.GetUserByEmail(email)
+	if err != nil {
+		// provision new member in THIS org
+		u, err = services.CreateUser(orgID, email, "", "member", "oidc")
+		if err != nil {
+			c.JSON(http.StatusInternalServerError, gin.H{"error": "provisioning failed"})
+			return
+		}
+	} else if u.OrgID != orgID {
+		c.JSON(http.StatusForbidden, gin.H{"error": "email belongs to a different organization"})
+		return
+	}
+
 	sessionID, err := SaveSession(ctx, &Session{
-		UserID: claims.Sub,
-		Email:  claims.Email,
-		Name:   claims.Name,
+		UserID: u.UserID, OrgID: u.OrgID, Role: u.Role, Email: u.Email, Name: claims.Name,
 	})
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": "session save failed"})
 		return
 	}
-
-	secure := c.Request.TLS != nil || c.GetHeader("X-Forwarded-Proto") == "https"
-	http.SetCookie(c.Writer, &http.Cookie{
-		Name:     sessionCookieName,
-		Value:    sessionID,
-		Path:     "/",
-		HttpOnly: true,
-		Secure:   secure,
-		SameSite: http.SameSiteLaxMode,
-		MaxAge:   int(sessionTTL.Seconds()),
-	})
-
-	frontendURL := os.Getenv("PUBLIC_HOST")
-	if frontendURL == "" {
-		frontendURL = "/"
-	}
-	c.Redirect(http.StatusFound, frontendURL)
+	_ = services.TouchLastLogin(u.UserID)
+	SetSessionCookie(c, sessionID)
+	c.Redirect(http.StatusFound, "/")
 }
 
 func HandleLogout(c *gin.Context) {
@@ -134,21 +166,3 @@ func HandleLogout(c *gin.Context) {
 	})
 	c.Redirect(http.StatusFound, "/")
 }
-
-func HandleMe(c *gin.Context) {
-	if !authEnabled {
-		c.JSON(http.StatusOK, gin.H{"auth_enabled": false})
-		return
-	}
-	cookie, err := c.Request.Cookie(sessionCookieName)
-	if err != nil {
-		c.JSON(http.StatusUnauthorized, gin.H{"error": "not authenticated"})
-		return
-	}
-	sess, err := GetSession(c.Request.Context(), cookie.Value)
-	if err != nil {
-		c.JSON(http.StatusUnauthorized, gin.H{"error": "session expired"})
-		return
-	}
-	c.JSON(http.StatusOK, sess)
-}

server/internal/auth/orghost.go

@@ -0,0 +1,80 @@
+package auth
+
+import (
+	"os"
+	"strings"
+	"sync"
+	"time"
+
+	"github.com/gin-gonic/gin"
+	"github.com/mrhid6/vantage/server/internal/models"
+	"github.com/mrhid6/vantage/server/internal/services"
+)
+
+type cachedOrg struct {
+	org *models.Org
+	at  time.Time
+}
+
+var (
+	orgCacheMu sync.Mutex
+	orgCache   = map[string]cachedOrg{}
+)
+
+const orgCacheTTL = 60 * time.Second
+
+// appRootLabel is the DNS label the app is deployed under, i.e. the "vantage"
+// in <slug>.vantage.<tld>. Deployments on another root must set APP_ROOT_LABEL
+// or every host resolves to no org, disabling the host/session mismatch guard.
+func appRootLabel() string {
+	if v := os.Getenv("APP_ROOT_LABEL"); v != "" {
+		return strings.ToLower(v)
+	}
+	return "vantage"
+}
+
+// hostSlug extracts the leftmost DNS label if the host is a subdomain of the
+// app root. Returns "" for the apex or an unknown host shape.
+func hostSlug(host string) string {
+	host = strings.ToLower(host)
+	if i := strings.IndexByte(host, ':'); i >= 0 {
+		host = host[:i]
+	}
+	root := appRootLabel()
+	// Expect <slug>.<root>.<...>; apex is <root>.<...>
+	parts := strings.Split(host, ".")
+	if len(parts) < 3 {
+		return ""
+	}
+	if parts[1] != root {
+		return ""
+	}
+	if parts[0] == root || parts[0] == "www" {
+		return ""
+	}
+	return parts[0]
+}
+
+func OrgFromHost(c *gin.Context) (*models.Org, bool) {
+	slug := hostSlug(c.Request.Host)
+	if slug == "" {
+		return nil, false
+	}
+	orgCacheMu.Lock()
+	if e, ok := orgCache[slug]; ok && time.Since(e.at) < orgCacheTTL {
+		orgCacheMu.Unlock()
+		return e.org, e.org != nil
+	}
+	orgCacheMu.Unlock()
+
+	org, err := services.GetOrgBySlug(slug)
+	if err != nil || org == nil {
+		// Never cache a miss: a just-bootstrapped org would otherwise 404 on its
+		// own subdomain for the rest of the TTL. Misses are cheap and rare.
+		return nil, false
+	}
+	orgCacheMu.Lock()
+	orgCache[slug] = cachedOrg{org: org, at: time.Now()}
+	orgCacheMu.Unlock()
+	return org, true
+}

server/internal/auth/session.go

@@ -17,6 +17,8 @@ const statePrefix = "km:state:"
 
 type Session struct {
 	UserID string `json:"user_id"`
+	OrgID  string `json:"org_id"`
+	Role   string `json:"role"`
 	Email  string `json:"email"`
 	Name   string `json:"name"`
 }
@@ -69,11 +71,14 @@ func DeleteSession(ctx context.Context, id string) error {
 	return rdb.Del(ctx, sessionPrefix+id).Err()
 }
 
-func SaveState(ctx context.Context, state string) error {
-	return rdb.Set(ctx, statePrefix+state, "1", 10*time.Minute).Err()
+func SaveStateOrg(ctx context.Context, state, orgID string) error {
+	return rdb.Set(ctx, statePrefix+state, orgID, 10*time.Minute).Err()
 }
 
-func ConsumeState(ctx context.Context, state string) bool {
-	n, err := rdb.Del(ctx, statePrefix+state).Result()
-	return err == nil && n > 0
+func ConsumeStateOrg(ctx context.Context, state string) (string, bool) {
+	orgID, err := rdb.GetDel(ctx, statePrefix+state).Result()
+	if err != nil || orgID == "" {
+		return "", false
+	}
+	return orgID, true
 }

server/internal/checker/checker.go

@@ -0,0 +1,226 @@
+// Package checker runs service checks (http/tcp/icmp/tls) and returns a uniform
+// Result. It has no dependency on models or pb so it can be duplicated verbatim
+// into the agent module (agent-run monitors) — callers map their own monitor
+// representation onto Spec.
+package checker
+
+import (
+	"context"
+	"crypto/tls"
+	"fmt"
+	"io"
+	"net"
+	"net/http"
+	"os"
+	"strings"
+	"time"
+)
+
+// Check types (mirror models.Monitor* constants).
+const (
+	TypeHTTP = "http"
+	TypeTCP  = "tcp"
+	TypeICMP = "icmp"
+	TypeTLS  = "tls"
+)
+
+// Spec is a self-contained description of a single check.
+type Spec struct {
+	Type           string
+	URL            string
+	Host           string
+	Port           int
+	Method         string
+	ExpectedStatus int
+	Keyword        string
+	TLSWarnDays    int
+	Insecure       bool // skip TLS certificate verification (HTTP checks)
+	TimeoutSec     int
+}
+
+// Result is the uniform outcome of running a check.
+type Result struct {
+	Up         bool
+	LatencyMs  int
+	Message    string
+	CertExpiry *time.Time
+}
+
+func (s Spec) timeout() time.Duration {
+	t := s.TimeoutSec
+	if t <= 0 || t > 10 {
+		t = 10
+	}
+	return time.Duration(t) * time.Second
+}
+
+// Run executes the check described by s.
+func Run(ctx context.Context, s Spec) Result {
+	switch s.Type {
+	case TypeHTTP:
+		return runHTTP(ctx, s)
+	case TypeTCP:
+		return runTCP(ctx, s)
+	case TypeICMP:
+		return runICMP(ctx, s)
+	case TypeTLS:
+		return runTLS(ctx, s)
+	default:
+		return Result{Message: "unknown check type: " + s.Type}
+	}
+}
+
+func runHTTP(ctx context.Context, s Spec) Result {
+	method := s.Method
+	if method == "" {
+		method = http.MethodGet
+	}
+	expect := s.ExpectedStatus
+	if expect == 0 {
+		expect = 200
+	}
+	client := &http.Client{Timeout: s.timeout()}
+	if s.Insecure {
+		client.Transport = &http.Transport{TLSClientConfig: &tls.Config{InsecureSkipVerify: true}} //nolint:gosec // opt-in per monitor
+	}
+	start := time.Now()
+	req, err := http.NewRequestWithContext(ctx, method, s.URL, nil)
+	if err != nil {
+		return Result{Message: err.Error()}
+	}
+	resp, err := client.Do(req)
+	if err != nil {
+		return Result{LatencyMs: msSince(start), Message: err.Error()}
+	}
+	defer resp.Body.Close()
+	res := Result{LatencyMs: msSince(start), Up: true}
+	if resp.TLS != nil && len(resp.TLS.PeerCertificates) > 0 {
+		exp := resp.TLS.PeerCertificates[0].NotAfter
+		res.CertExpiry = &exp
+	}
+	if resp.StatusCode != expect {
+		return Result{LatencyMs: res.LatencyMs, CertExpiry: res.CertExpiry, Message: fmt.Sprintf("status %d (want %d)", resp.StatusCode, expect)}
+	}
+	if s.Keyword != "" {
+		body, _ := io.ReadAll(io.LimitReader(resp.Body, 1<<20))
+		if !strings.Contains(string(body), s.Keyword) {
+			return Result{LatencyMs: res.LatencyMs, CertExpiry: res.CertExpiry, Message: "keyword not found"}
+		}
+	}
+	return res
+}
+
+func runTCP(ctx context.Context, s Spec) Result {
+	addr := net.JoinHostPort(s.Host, fmt.Sprint(s.Port))
+	start := time.Now()
+	d := net.Dialer{Timeout: s.timeout()}
+	conn, err := d.DialContext(ctx, "tcp", addr)
+	if err != nil {
+		return Result{LatencyMs: msSince(start), Message: err.Error()}
+	}
+	conn.Close()
+	return Result{Up: true, LatencyMs: msSince(start)}
+}
+
+func runTLS(ctx context.Context, s Spec) Result {
+	port := s.Port
+	if port == 0 {
+		port = 443
+	}
+	addr := net.JoinHostPort(s.Host, fmt.Sprint(port))
+	start := time.Now()
+	d := net.Dialer{Timeout: s.timeout()}
+	conn, err := tls.DialWithDialer(&d, "tcp", addr, &tls.Config{ServerName: s.Host})
+	if err != nil {
+		return Result{LatencyMs: msSince(start), Message: err.Error()}
+	}
+	defer conn.Close()
+	certs := conn.ConnectionState().PeerCertificates
+	if len(certs) == 0 {
+		return Result{LatencyMs: msSince(start), Message: "no peer certificate"}
+	}
+	exp := certs[0].NotAfter
+	res := Result{LatencyMs: msSince(start), CertExpiry: &exp}
+	warn := s.TLSWarnDays
+	if warn <= 0 {
+		warn = 14
+	}
+	remaining := time.Until(exp)
+	if remaining <= 0 {
+		res.Message = "certificate expired"
+		return res
+	}
+	if remaining <= time.Duration(warn)*24*time.Hour {
+		res.Message = fmt.Sprintf("certificate expires in %d days", int(remaining.Hours()/24))
+		return res
+	}
+	res.Up = true
+	return res
+}
+
+func msSince(t time.Time) int { return int(time.Since(t).Milliseconds()) }
+
+// runICMP sends a single ICMP echo request and waits for the reply. Requires
+// raw-socket privileges (the agent and server run as root). Returns down with a
+// descriptive message when the socket cannot be opened or no reply arrives.
+func runICMP(ctx context.Context, s Spec) Result {
+	dst, err := net.ResolveIPAddr("ip4", s.Host)
+	if err != nil {
+		return Result{Message: err.Error()}
+	}
+	conn, err := net.ListenPacket("ip4:icmp", "0.0.0.0")
+	if err != nil {
+		return Result{Message: "icmp socket: " + err.Error()}
+	}
+	defer conn.Close()
+
+	id := os.Getpid() & 0xffff
+	pkt := icmpEcho(id, 1)
+	deadline := time.Now().Add(s.timeout())
+	if d, ok := ctx.Deadline(); ok && d.Before(deadline) {
+		deadline = d
+	}
+	_ = conn.SetDeadline(deadline)
+
+	start := time.Now()
+	if _, err := conn.WriteTo(pkt, dst); err != nil {
+		return Result{Message: err.Error()}
+	}
+	reply := make([]byte, 1500)
+	for {
+		n, peer, err := conn.ReadFrom(reply)
+		if err != nil {
+			return Result{LatencyMs: msSince(start), Message: "no reply"}
+		}
+		// Skip the IPv4 header (20 bytes) to reach the ICMP message.
+		if n < 28 || peer.String() != dst.String() {
+			continue
+		}
+		if reply[20] == 0 { // ICMP echo reply type
+			return Result{Up: true, LatencyMs: msSince(start)}
+		}
+	}
+}
+
+func icmpEcho(id, seq int) []byte {
+	// Type(8)=echo request, Code=0, Checksum, ID, Seq, no payload.
+	b := []byte{8, 0, 0, 0, byte(id >> 8), byte(id), byte(seq >> 8), byte(seq)}
+	cs := icmpChecksum(b)
+	b[2] = byte(cs >> 8)
+	b[3] = byte(cs)
+	return b
+}
+
+func icmpChecksum(b []byte) uint16 {
+	var sum uint32
+	for i := 0; i < len(b)-1; i += 2 {
+		sum += uint32(b[i])<<8 | uint32(b[i+1])
+	}
+	if len(b)%2 == 1 {
+		sum += uint32(b[len(b)-1]) << 8
+	}
+	for sum>>16 != 0 {
+		sum = (sum & 0xffff) + (sum >> 16)
+	}
+	return ^uint16(sum)
+}

server/internal/grpc/pb/vantage.pb.go

@@ -63,6 +63,75 @@ type ReportUpdatesRequest struct {
 
 type ReportUpdatesResponse struct{}
 
+// Inventory report message types
+
+type CPUReport struct {
+	Model    string  `json:"model,omitempty"`
+	Cores    int     `json:"cores,omitempty"`
+	UsagePct float64 `json:"usage_pct"`
+	Load1    float64 `json:"load1,omitempty"`
+}
+type MemReport struct {
+	TotalBytes uint64 `json:"total_bytes"`
+	UsedBytes  uint64 `json:"used_bytes"`
+}
+type PartitionReport struct {
+	Device     string `json:"device"`
+	Mountpoint string `json:"mountpoint"`
+	Fstype     string `json:"fstype,omitempty"`
+	TotalBytes uint64 `json:"total_bytes"`
+	UsedBytes  uint64 `json:"used_bytes"`
+}
+type InventoryReport struct {
+	ServerId      string            `json:"server_id"`
+	AgentToken    string            `json:"agent_token"`
+	IncludeStatic bool              `json:"include_static"`
+	CPU           *CPUReport        `json:"cpu,omitempty"`
+	Memory        *MemReport        `json:"memory,omitempty"`
+	SwapTotal     uint64            `json:"swap_total"`
+	SwapUsed      uint64            `json:"swap_used"`
+	Partitions    []PartitionReport `json:"partitions,omitempty"`
+	Kernel        string            `json:"kernel,omitempty"`
+}
+type InventoryReportResponse struct{}
+
+// Monitor sync / check report message types
+
+type MonitorSpec struct {
+	MonitorId      string `json:"monitor_id"`
+	Type           string `json:"type"`
+	URL            string `json:"url,omitempty"`
+	Host           string `json:"host,omitempty"`
+	Port           int    `json:"port,omitempty"`
+	Method         string `json:"method,omitempty"`
+	ExpectedStatus int    `json:"expected_status,omitempty"`
+	Keyword        string `json:"keyword,omitempty"`
+	TLSWarnDays    int    `json:"tls_warn_days,omitempty"`
+	Insecure       bool   `json:"insecure,omitempty"`
+	IntervalSec    int    `json:"interval_sec"`
+	Retries        int    `json:"retries"`
+}
+type SyncMonitorsRequest struct {
+	ServerId   string `json:"server_id"`
+	AgentToken string `json:"agent_token"`
+}
+type SyncMonitorsResponse struct {
+	Monitors []MonitorSpec `json:"monitors,omitempty"`
+}
+type CheckResult struct {
+	MonitorId     string `json:"monitor_id"`
+	Up            bool   `json:"up"`
+	LatencyMs     int    `json:"latency_ms"`
+	Message       string `json:"message,omitempty"`
+	CertExpiryUnix int64 `json:"cert_expiry_unix,omitempty"`
+}
+type ReportChecksRequest struct {
+	ServerId   string        `json:"server_id"`
+	AgentToken string        `json:"agent_token"`
+	Results    []CheckResult `json:"results,omitempty"`
+}
+type ReportChecksResponse struct{}
+
 type ApplyUpdatesCmd struct{}
 
 type ServerCommand struct {
@@ -195,6 +264,9 @@ type VantageServer interface {
 	SyncKeys(context.Context, *SyncRequest) (*SyncResponse, error)
 	UploadGeneratedKey(context.Context, *UploadKeyRequest) (*UploadKeyResponse, error)
 	ReportUpdates(context.Context, *ReportUpdatesRequest) (*ReportUpdatesResponse, error)
+	ReportInventory(context.Context, *InventoryReport) (*InventoryReportResponse, error)
+	SyncMonitors(context.Context, *SyncMonitorsRequest) (*SyncMonitorsResponse, error)
+	ReportChecks(context.Context, *ReportChecksRequest) (*ReportChecksResponse, error)
 	CommandStream(Vantage_CommandStreamServer) error
 }
 
@@ -216,6 +288,18 @@ func (UnimplementedVantageServer) ReportUpdates(context.Context, *ReportUpdatesR
 	return nil, status.Errorf(codes.Unimplemented, "method ReportUpdates not implemented")
 }
 
+func (UnimplementedVantageServer) ReportInventory(context.Context, *InventoryReport) (*InventoryReportResponse, error) {
+	return nil, status.Errorf(codes.Unimplemented, "method ReportInventory not implemented")
+}
+
+func (UnimplementedVantageServer) SyncMonitors(context.Context, *SyncMonitorsRequest) (*SyncMonitorsResponse, error) {
+	return nil, status.Errorf(codes.Unimplemented, "method SyncMonitors not implemented")
+}
+
+func (UnimplementedVantageServer) ReportChecks(context.Context, *ReportChecksRequest) (*ReportChecksResponse, error) {
+	return nil, status.Errorf(codes.Unimplemented, "method ReportChecks not implemented")
+}
+
 func (UnimplementedVantageServer) CommandStream(Vantage_CommandStreamServer) error {
 	return status.Errorf(codes.Unimplemented, "method CommandStream not implemented")
 }
@@ -227,6 +311,9 @@ type VantageClient interface {
 	SyncKeys(ctx context.Context, in *SyncRequest, opts ...grpc.CallOption) (*SyncResponse, error)
 	UploadGeneratedKey(ctx context.Context, in *UploadKeyRequest, opts ...grpc.CallOption) (*UploadKeyResponse, error)
 	ReportUpdates(ctx context.Context, in *ReportUpdatesRequest, opts ...grpc.CallOption) (*ReportUpdatesResponse, error)
+	ReportInventory(ctx context.Context, in *InventoryReport, opts ...grpc.CallOption) (*InventoryReportResponse, error)
+	SyncMonitors(ctx context.Context, in *SyncMonitorsRequest, opts ...grpc.CallOption) (*SyncMonitorsResponse, error)
+	ReportChecks(ctx context.Context, in *ReportChecksRequest, opts ...grpc.CallOption) (*ReportChecksResponse, error)
 	CommandStream(ctx context.Context, opts ...grpc.CallOption) (Vantage_CommandStreamClient, error)
 }
 
@@ -270,6 +357,30 @@ func (c *keyManagerClient) ReportUpdates(ctx context.Context, in *ReportUpdatesR
 	return out, nil
 }
 
+func (c *keyManagerClient) ReportInventory(ctx context.Context, in *InventoryReport, opts ...grpc.CallOption) (*InventoryReportResponse, error) {
+	out := new(InventoryReportResponse)
+	if err := c.cc.Invoke(ctx, "/vantage.v1.Vantage/ReportInventory", in, out, opts...); err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
+func (c *keyManagerClient) SyncMonitors(ctx context.Context, in *SyncMonitorsRequest, opts ...grpc.CallOption) (*SyncMonitorsResponse, error) {
+	out := new(SyncMonitorsResponse)
+	if err := c.cc.Invoke(ctx, "/vantage.v1.Vantage/SyncMonitors", in, out, opts...); err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
+func (c *keyManagerClient) ReportChecks(ctx context.Context, in *ReportChecksRequest, opts ...grpc.CallOption) (*ReportChecksResponse, error) {
+	out := new(ReportChecksResponse)
+	if err := c.cc.Invoke(ctx, "/vantage.v1.Vantage/ReportChecks", in, out, opts...); err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
 func (c *keyManagerClient) CommandStream(ctx context.Context, opts ...grpc.CallOption) (Vantage_CommandStreamClient, error) {
 	stream, err := c.cc.NewStream(ctx, &Vantage_ServiceDesc.Streams[0], "/vantage.v1.Vantage/CommandStream", opts...)
 	if err != nil {
@@ -292,6 +403,9 @@ var Vantage_ServiceDesc = grpc.ServiceDesc{
 		{MethodName: "SyncKeys", Handler: _Vantage_SyncKeys_Handler},
 		{MethodName: "UploadGeneratedKey", Handler: _Vantage_UploadGeneratedKey_Handler},
 		{MethodName: "ReportUpdates", Handler: _Vantage_ReportUpdates_Handler},
+		{MethodName: "ReportInventory", Handler: _Vantage_ReportInventory_Handler},
+		{MethodName: "SyncMonitors", Handler: _Vantage_SyncMonitors_Handler},
+		{MethodName: "ReportChecks", Handler: _Vantage_ReportChecks_Handler},
 	},
 	Streams: []grpc.StreamDesc{
 		{
@@ -364,6 +478,51 @@ func _Vantage_ReportUpdates_Handler(srv interface{}, ctx context.Context, dec fu
 	return interceptor(ctx, in, info, handler)
 }
 
+func _Vantage_ReportInventory_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
+	in := new(InventoryReport)
+	if err := dec(in); err != nil {
+		return nil, err
+	}
+	if interceptor == nil {
+		return srv.(VantageServer).ReportInventory(ctx, in)
+	}
+	info := &grpc.UnaryServerInfo{Server: srv, FullMethod: "/vantage.v1.Vantage/ReportInventory"}
+	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
+		return srv.(VantageServer).ReportInventory(ctx, req.(*InventoryReport))
+	}
+	return interceptor(ctx, in, info, handler)
+}
+
+func _Vantage_SyncMonitors_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
+	in := new(SyncMonitorsRequest)
+	if err := dec(in); err != nil {
+		return nil, err
+	}
+	if interceptor == nil {
+		return srv.(VantageServer).SyncMonitors(ctx, in)
+	}
+	info := &grpc.UnaryServerInfo{Server: srv, FullMethod: "/vantage.v1.Vantage/SyncMonitors"}
+	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
+		return srv.(VantageServer).SyncMonitors(ctx, req.(*SyncMonitorsRequest))
+	}
+	return interceptor(ctx, in, info, handler)
+}
+
+func _Vantage_ReportChecks_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
+	in := new(ReportChecksRequest)
+	if err := dec(in); err != nil {
+		return nil, err
+	}
+	if interceptor == nil {
+		return srv.(VantageServer).ReportChecks(ctx, in)
+	}
+	info := &grpc.UnaryServerInfo{Server: srv, FullMethod: "/vantage.v1.Vantage/ReportChecks"}
+	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
+		return srv.(VantageServer).ReportChecks(ctx, req.(*ReportChecksRequest))
+	}
+	return interceptor(ctx, in, info, handler)
+}
+
 func _Vantage_CommandStream_Handler(srv interface{}, stream grpc.ServerStream) error {
 	return srv.(VantageServer).CommandStream(&keyManagerCommandStreamServer{stream})
 }

server/internal/grpc/server.go

@@ -7,6 +7,7 @@ import (
 	"net"
 	"time"
 
+	"github.com/mrhid6/vantage/server/internal/checker"
 	"github.com/mrhid6/vantage/server/internal/grpc/pb"
 	"github.com/mrhid6/vantage/server/internal/models"
 	"github.com/mrhid6/vantage/server/internal/services"
@@ -62,13 +63,13 @@ func (s *vantageServer) UploadGeneratedKey(ctx context.Context, req *pb.UploadKe
 	}
 
 	// Agent-generated keys carry no passphrase over the wire (proto has no field).
-	key, err := services.CreateKey(req.Label, req.PublicKey, "generated", srv.ServerID, req.PrivateKey, "")
+	key, err := services.CreateKey(srv.OrgID, req.Label, req.PublicKey, "generated", srv.ServerID, req.PrivateKey, "")
 	if err != nil {
 		return nil, status.Errorf(codes.Internal, "failed to store key: %v", err)
 	}
 
 	// Auto-assign to the generating server
-	if _, err := services.AssignKey(key.KeyID, srv.ServerID); err != nil {
+	if _, err := services.AssignKey(srv.OrgID, key.KeyID, srv.ServerID); err != nil {
 		log.Printf("failed to auto-assign generated key: %v", err)
 	}
 
@@ -95,6 +96,66 @@ func (s *vantageServer) ReportUpdates(ctx context.Context, req *pb.ReportUpdates
 	return &pb.ReportUpdatesResponse{}, nil
 }
 
+func (s *vantageServer) ReportInventory(ctx context.Context, req *pb.InventoryReport) (*pb.InventoryReportResponse, error) {
+	srv, err := services.ValidateAgentToken(req.ServerId, req.AgentToken)
+	if err != nil {
+		return nil, status.Errorf(codes.Unauthenticated, "invalid agent token")
+	}
+	if err := services.StoreInventory(srv.ServerID, req); err != nil {
+		log.Printf("store inventory for %s: %v", srv.ServerID, err)
+	}
+	return &pb.InventoryReportResponse{}, nil
+}
+
+func (s *vantageServer) SyncMonitors(ctx context.Context, req *pb.SyncMonitorsRequest) (*pb.SyncMonitorsResponse, error) {
+	srv, err := services.ValidateAgentToken(req.ServerId, req.AgentToken)
+	if err != nil {
+		return nil, status.Errorf(codes.Unauthenticated, "invalid agent token")
+	}
+	monitors, err := services.ListMonitorsForRunner(srv.OrgID, srv.ServerID)
+	if err != nil {
+		return nil, status.Errorf(codes.Internal, "list monitors")
+	}
+	specs := make([]pb.MonitorSpec, 0, len(monitors))
+	for _, m := range monitors {
+		specs = append(specs, pb.MonitorSpec{
+			MonitorId:      m.MonitorID,
+			Type:           m.Type,
+			URL:            m.Target.URL,
+			Host:           m.Target.Host,
+			Port:           m.Target.Port,
+			Method:         m.Target.Method,
+			ExpectedStatus: m.Target.ExpectedStatus,
+			Keyword:        m.Target.Keyword,
+			TLSWarnDays:    m.Target.TLSWarnDays,
+			Insecure:       m.Target.Insecure,
+			IntervalSec:    m.IntervalSec,
+			Retries:        m.Retries,
+		})
+	}
+	return &pb.SyncMonitorsResponse{Monitors: specs}, nil
+}
+
+func (s *vantageServer) ReportChecks(ctx context.Context, req *pb.ReportChecksRequest) (*pb.ReportChecksResponse, error) {
+	srv, err := services.ValidateAgentToken(req.ServerId, req.AgentToken)
+	if err != nil {
+		return nil, status.Errorf(codes.Unauthenticated, "invalid agent token")
+	}
+	for _, r := range req.Results {
+		res := checker.Result{Up: r.Up, LatencyMs: r.LatencyMs, Message: r.Message}
+		if r.CertExpiryUnix > 0 {
+			t := time.Unix(r.CertExpiryUnix, 0)
+			res.CertExpiry = &t
+		}
+		// A rejected monitor (wrong org, or not run by this agent) is skipped,
+		// not fatal — the rest of the batch is still legitimate.
+		if err := services.IngestResult(srv.OrgID, srv.ServerID, r.MonitorId, res); err != nil {
+			log.Printf("ingest check %s: %v", r.MonitorId, err)
+		}
+	}
+	return &pb.ReportChecksResponse{}, nil
+}
+
 func (s *vantageServer) CommandStream(stream pb.Vantage_CommandStreamServer) error {
 	// First message authenticates the agent and signals readiness.
 	msg, err := stream.Recv()

server/internal/models/assignment.go

@@ -8,8 +8,9 @@ import (
 
 type Assignment struct {
 	ID         bson.ObjectID `bson:"_id,omitempty" json:"_id,omitempty"`
-	KeyID      string             `bson:"key_id" json:"key_id"`
-	ServerID   string             `bson:"server_id" json:"server_id"`
-	AssignedAt time.Time          `bson:"assigned_at" json:"assigned_at"`
-	RevokedAt  *time.Time         `bson:"revoked_at,omitempty" json:"revoked_at,omitempty"`
+	OrgID      string        `bson:"org_id" json:"org_id"`
+	KeyID      string        `bson:"key_id" json:"key_id"`
+	ServerID   string        `bson:"server_id" json:"server_id"`
+	AssignedAt time.Time     `bson:"assigned_at" json:"assigned_at"`
+	RevokedAt  *time.Time    `bson:"revoked_at,omitempty" json:"revoked_at,omitempty"`
 }

server/internal/models/audit.go

@@ -8,6 +8,7 @@ import (
 
 type AuditEvent struct {
 	ID        bson.ObjectID `bson:"_id,omitempty" json:"id"`
+	OrgID     string        `bson:"org_id" json:"org_id"`
 	EventType string        `bson:"event_type" json:"event_type"`
 	Actor     string        `bson:"actor" json:"actor"`
 	ServerID  string        `bson:"server_id,omitempty" json:"server_id,omitempty"`

server/internal/models/channel.go

@@ -0,0 +1,30 @@
+package models
+
+import (
+	"time"
+
+	"go.mongodb.org/mongo-driver/v2/bson"
+)
+
+// Notification channel types.
+const (
+	ChannelWebhook  = "webhook"
+	ChannelSMTP     = "smtp"
+	ChannelDiscord  = "discord"
+	ChannelSlack    = "slack"
+	ChannelTelegram = "telegram"
+)
+
+// NotificationChannel is an outbound alert destination. Config holds
+// type-specific settings (e.g. url; or smtp host/port/username/password/from/to;
+// or telegram token/chat_id).
+type NotificationChannel struct {
+	ID        bson.ObjectID     `bson:"_id,omitempty" json:"_id,omitempty"`
+	OrgID     string            `bson:"org_id" json:"org_id"`
+	ChannelID string            `bson:"channel_id" json:"channel_id"`
+	Name      string            `bson:"name" json:"name"`
+	Type      string            `bson:"type" json:"type"`
+	Config    map[string]string `bson:"config" json:"config"`
+	Enabled   bool              `bson:"enabled" json:"enabled"`
+	CreatedAt time.Time         `bson:"created_at" json:"created_at"`
+}

server/internal/models/console_session.go

@@ -8,6 +8,7 @@ import (
 
 type ConsoleSession struct {
 	ID        bson.ObjectID `bson:"_id,omitempty" json:"_id,omitempty"`
+	OrgID     string        `bson:"org_id" json:"org_id"`
 	SessionID string        `bson:"session_id" json:"session_id"`
 	ServerID  string        `bson:"server_id" json:"server_id"`
 	Protocol  string        `bson:"protocol" json:"protocol"` // ssh | rdp | vnc

server/internal/models/key.go

@@ -8,6 +8,7 @@ import (
 
 type Key struct {
 	ID                  bson.ObjectID `bson:"_id,omitempty" json:"_id,omitempty"`
+	OrgID               string        `bson:"org_id" json:"org_id"`
 	KeyID               string        `bson:"key_id" json:"key_id"`
 	Label               string        `bson:"label" json:"label"`
 	PublicKey           string        `bson:"public_key" json:"public_key"`

server/internal/models/monitor.go

@@ -0,0 +1,81 @@
+package models
+
+import (
+	"time"
+
+	"go.mongodb.org/mongo-driver/v2/bson"
+)
+
+// Monitor check types.
+const (
+	MonitorHTTP = "http"
+	MonitorTCP  = "tcp"
+	MonitorICMP = "icmp"
+	MonitorTLS  = "tls"
+)
+
+// Monitor status values.
+const (
+	StatusUp      = "up"
+	StatusDown    = "down"
+	StatusPending = "pending"
+)
+
+// RunnerServer is the reserved Runner value for server-run monitors. Any other
+// value is treated as a server_id whose agent runs the check locally.
+const RunnerServer = "server"
+
+type MonitorTarget struct {
+	URL            string `bson:"url,omitempty" json:"url,omitempty"`
+	Host           string `bson:"host,omitempty" json:"host,omitempty"`
+	Port           int    `bson:"port,omitempty" json:"port,omitempty"`
+	Method         string `bson:"method,omitempty" json:"method,omitempty"`
+	ExpectedStatus int    `bson:"expected_status,omitempty" json:"expected_status,omitempty"`
+	Keyword        string `bson:"keyword,omitempty" json:"keyword,omitempty"`
+	TLSWarnDays    int    `bson:"tls_warn_days,omitempty" json:"tls_warn_days,omitempty"`
+	Insecure       bool   `bson:"insecure,omitempty" json:"insecure,omitempty"` // skip TLS cert verification (HTTP monitors)
+}
+
+type MonitorState struct {
+	Status         string     `bson:"status" json:"status"` // up|down|pending
+	LastCheckAt    *time.Time `bson:"last_check_at,omitempty" json:"last_check_at,omitempty"`
+	LatencyMs      int        `bson:"latency_ms" json:"latency_ms"`
+	Message        string     `bson:"message,omitempty" json:"message,omitempty"`
+	CertExpiryAt   *time.Time `bson:"cert_expiry_at,omitempty" json:"cert_expiry_at,omitempty"`
+	Fails          int        `bson:"fails" json:"fails"` // consecutive failures
+	LastNotifiedAt *time.Time `bson:"last_notified_at,omitempty" json:"last_notified_at,omitempty"`
+}
+
+type Monitor struct {
+	ID          bson.ObjectID `bson:"_id,omitempty" json:"_id,omitempty"`
+	OrgID       string        `bson:"org_id" json:"org_id"`
+	MonitorID   string        `bson:"monitor_id" json:"monitor_id"`
+	Name        string        `bson:"name" json:"name"`
+	Type        string        `bson:"type" json:"type"` // http|tcp|icmp|tls
+	Target      MonitorTarget `bson:"target" json:"target"`
+	IntervalSec int           `bson:"interval_sec" json:"interval_sec"`
+	Runner      string        `bson:"runner" json:"runner"`   // "server" or a server_id
+	Retries     int           `bson:"retries" json:"retries"` // consecutive fails before down
+	Enabled     bool          `bson:"enabled" json:"enabled"`
+	ChannelIDs  []string      `bson:"channel_ids,omitempty" json:"channel_ids,omitempty"`
+	State       MonitorState  `bson:"state" json:"state"`
+	CreatedAt   time.Time     `bson:"created_at" json:"created_at"`
+}
+
+type Incident struct {
+	OrgID      string     `bson:"org_id" json:"org_id"`
+	IncidentID string     `bson:"incident_id" json:"incident_id"`
+	MonitorID  string     `bson:"monitor_id" json:"monitor_id"`
+	StartedAt  time.Time  `bson:"started_at" json:"started_at"`
+	ResolvedAt *time.Time `bson:"resolved_at,omitempty" json:"resolved_at,omitempty"`
+	Cause      string     `bson:"cause,omitempty" json:"cause,omitempty"`
+}
+
+type Rollup struct {
+	OrgID       string    `bson:"org_id" json:"org_id"`
+	MonitorID   string    `bson:"monitor_id" json:"monitor_id"`
+	PeriodStart time.Time `bson:"period_start" json:"period_start"` // hour bucket
+	Checks      int       `bson:"checks" json:"checks"`
+	UpCount     int       `bson:"up_count" json:"up_count"`
+	SumLatency  int64     `bson:"sum_latency" json:"sum_latency"`
+}

server/internal/models/org.go

@@ -0,0 +1,15 @@
+package models
+
+import (
+	"time"
+
+	"go.mongodb.org/mongo-driver/v2/bson"
+)
+
+type Org struct {
+	ID        bson.ObjectID `bson:"_id,omitempty" json:"_id,omitempty"`
+	OrgID     string        `bson:"org_id" json:"org_id"`
+	Name      string        `bson:"name" json:"name"`
+	Slug      string        `bson:"slug" json:"slug"`
+	CreatedAt time.Time     `bson:"created_at" json:"created_at"`
+}

server/internal/models/org_oidc.go

@@ -0,0 +1,17 @@
+package models
+
+import (
+	"time"
+
+	"go.mongodb.org/mongo-driver/v2/bson"
+)
+
+type OrgOIDC struct {
+	ID              bson.ObjectID `bson:"_id,omitempty" json:"_id,omitempty"`
+	OrgID           string        `bson:"org_id" json:"org_id"`
+	Issuer          string        `bson:"issuer" json:"issuer"`
+	ClientID        string        `bson:"client_id" json:"client_id"`
+	ClientSecretEnc string        `bson:"client_secret_enc,omitempty" json:"-"`
+	Enabled         bool          `bson:"enabled" json:"enabled"`
+	UpdatedAt       time.Time     `bson:"updated_at" json:"updated_at"`
+}

server/internal/models/secret.go

@@ -10,6 +10,7 @@ import (
 // encrypted (AES-256-GCM) and is never serialized to JSON.
 type Secret struct {
 	ID             bson.ObjectID `bson:"_id,omitempty" json:"-"`
+	OrgID          string        `bson:"org_id" json:"org_id"`
 	Group          string        `bson:"group" json:"group"`
 	Key            string        `bson:"key" json:"key"`
 	EncryptedValue string        `bson:"encrypted_value" json:"-"`

server/internal/models/server.go

@@ -12,23 +12,56 @@ type PackageUpdate struct {
 	NewVersion     string `bson:"new_version" json:"new_version"`
 }
 
-type Server struct {
-	ID               bson.ObjectID  `bson:"_id,omitempty" json:"_id,omitempty"`
-	ServerID         string         `bson:"server_id" json:"server_id"`
-	Hostname         string         `bson:"hostname" json:"hostname"`
-	IPAddress        string         `bson:"ip_address" json:"ip_address"`
-	OSInfo           string         `bson:"os_info" json:"os_info"`
-	OSType           string         `bson:"os_type,omitempty" json:"os_type,omitempty"`
-	ConsoleProtocols []string       `bson:"console_protocols,omitempty" json:"console_protocols,omitempty"`
-	SSHPort          int            `bson:"ssh_port,omitempty" json:"ssh_port,omitempty"`
-	RDPPort          int            `bson:"rdp_port,omitempty" json:"rdp_port,omitempty"`
-	PreRegToken      string         `bson:"pre_reg_token,omitempty" json:"pre_reg_token,omitempty"`
-	PreRegExpires    *time.Time     `bson:"pre_reg_expires,omitempty" json:"pre_reg_expires,omitempty"`
-	AgentTokenHash   string         `bson:"agent_token_hash,omitempty" json:"-"`
-	Status           string         `bson:"status" json:"status"`
-	AgentVersion     string         `bson:"agent_version,omitempty" json:"agent_version,omitempty"`
-	LastSeen         *time.Time     `bson:"last_seen,omitempty" json:"last_seen,omitempty"`
-	AvailableUpdates []PackageUpdate `bson:"available_updates,omitempty" json:"available_updates,omitempty"`
-	UpdatesCheckedAt *time.Time     `bson:"updates_checked_at,omitempty" json:"updates_checked_at,omitempty"`
-	CreatedAt        time.Time      `bson:"created_at" json:"created_at"`
+type CPUInfo struct {
+	Model    string  `bson:"model,omitempty" json:"model,omitempty"`
+	Cores    int     `bson:"cores,omitempty" json:"cores,omitempty"`
+	UsagePct float64 `bson:"usage_pct" json:"usage_pct"`
+	Load1    float64 `bson:"load1,omitempty" json:"load1,omitempty"`
+}
+
+type MemInfo struct {
+	TotalBytes uint64 `bson:"total_bytes" json:"total_bytes"`
+	UsedBytes  uint64 `bson:"used_bytes" json:"used_bytes"`
+}
+
+type Partition struct {
+	Device     string `bson:"device" json:"device"`
+	Mountpoint string `bson:"mountpoint" json:"mountpoint"`
+	Fstype     string `bson:"fstype,omitempty" json:"fstype,omitempty"`
+	TotalBytes uint64 `bson:"total_bytes" json:"total_bytes"`
+	UsedBytes  uint64 `bson:"used_bytes" json:"used_bytes"`
+}
+
+type Inventory struct {
+	CPU            CPUInfo     `bson:"cpu" json:"cpu"`
+	Memory         MemInfo     `bson:"memory" json:"memory"`
+	SwapTotalBytes uint64      `bson:"swap_total_bytes" json:"swap_total_bytes"`
+	SwapUsedBytes  uint64      `bson:"swap_used_bytes" json:"swap_used_bytes"`
+	Partitions     []Partition `bson:"partitions,omitempty" json:"partitions,omitempty"`
+	Kernel         string      `bson:"kernel,omitempty" json:"kernel,omitempty"`
+	MetricsAt      *time.Time  `bson:"metrics_at,omitempty" json:"metrics_at,omitempty"`
+	StaticAt       *time.Time  `bson:"static_at,omitempty" json:"static_at,omitempty"`
+}
+
+type Server struct {
+	ID               bson.ObjectID   `bson:"_id,omitempty" json:"_id,omitempty"`
+	OrgID            string          `bson:"org_id" json:"org_id"`
+	ServerID         string          `bson:"server_id" json:"server_id"`
+	Hostname         string          `bson:"hostname" json:"hostname"`
+	IPAddress        string          `bson:"ip_address" json:"ip_address"`
+	OSInfo           string          `bson:"os_info" json:"os_info"`
+	OSType           string          `bson:"os_type,omitempty" json:"os_type,omitempty"`
+	ConsoleProtocols []string        `bson:"console_protocols,omitempty" json:"console_protocols,omitempty"`
+	SSHPort          int             `bson:"ssh_port,omitempty" json:"ssh_port,omitempty"`
+	RDPPort          int             `bson:"rdp_port,omitempty" json:"rdp_port,omitempty"`
+	PreRegToken      string          `bson:"pre_reg_token,omitempty" json:"pre_reg_token,omitempty"`
+	PreRegExpires    *time.Time      `bson:"pre_reg_expires,omitempty" json:"pre_reg_expires,omitempty"`
+	AgentTokenHash   string          `bson:"agent_token_hash,omitempty" json:"-"`
+	Status           string          `bson:"status" json:"status"`
+	AgentVersion     string          `bson:"agent_version,omitempty" json:"agent_version,omitempty"`
+	LastSeen         *time.Time      `bson:"last_seen,omitempty" json:"last_seen,omitempty"`
+	AvailableUpdates []PackageUpdate `bson:"available_updates,omitempty" json:"available_updates,omitempty"`
+	UpdatesCheckedAt *time.Time      `bson:"updates_checked_at,omitempty" json:"updates_checked_at,omitempty"`
+	Inventory        *Inventory      `bson:"inventory,omitempty" json:"inventory,omitempty"`
+	CreatedAt        time.Time       `bson:"created_at" json:"created_at"`
 }

server/internal/models/settings.go

@@ -33,6 +33,7 @@ type SecretsSettings struct {
 
 type Settings struct {
 	ID      bson.ObjectID   `bson:"_id,omitempty" json:"-"`
+	OrgID   string          `bson:"org_id" json:"org_id"`
 	Alerts  AlertSettings   `bson:"alerts" json:"alerts"`
 	Email   EmailSettings   `bson:"email" json:"email"`
 	Secrets SecretsSettings `bson:"secrets" json:"secrets"`

server/internal/models/user.go

@@ -0,0 +1,35 @@
+package models
+
+import (
+	"time"
+
+	"go.mongodb.org/mongo-driver/v2/bson"
+)
+
+// Org membership roles. These are the only values ever written to User.Role;
+// anything arriving from a client must be checked with ValidRole first.
+const (
+	RoleOwner  = "owner"
+	RoleAdmin  = "admin"
+	RoleMember = "member"
+)
+
+func ValidRole(role string) bool {
+	switch role {
+	case RoleOwner, RoleAdmin, RoleMember:
+		return true
+	}
+	return false
+}
+
+type User struct {
+	ID           bson.ObjectID `bson:"_id,omitempty" json:"_id,omitempty"`
+	UserID       string        `bson:"user_id" json:"user_id"`
+	OrgID        string        `bson:"org_id" json:"org_id"`
+	Email        string        `bson:"email" json:"email"`
+	PasswordHash string        `bson:"password_hash,omitempty" json:"-"`
+	Role         string        `bson:"role" json:"role"`               // owner|admin|member
+	AuthSource   string        `bson:"auth_source" json:"auth_source"` // local|oidc
+	CreatedAt    time.Time     `bson:"created_at" json:"created_at"`
+	LastLogin    *time.Time    `bson:"last_login,omitempty" json:"last_login,omitempty"`
+}

server/internal/models/workflow.go

@@ -14,6 +14,7 @@ type InputParam struct {
 
 type WorkflowStep struct {
 	ID              bson.ObjectID `bson:"_id,omitempty" json:"-"`
+	OrgID           string        `bson:"org_id" json:"org_id"`
 	StepID          string        `bson:"step_id" json:"step_id"`
 	Name            string        `bson:"name" json:"name"`
 	Description     string        `bson:"description" json:"description"`
@@ -22,12 +23,15 @@ type WorkflowStep struct {
 	DeclaredOutputs []string      `bson:"declared_outputs" json:"declared_outputs"`
 	DeclaredInputs  []InputParam  `bson:"declared_inputs" json:"declared_inputs"`
 	SecretRefs      []string      `bson:"secret_refs" json:"secret_refs"`
+	Source          string        `bson:"source" json:"source"` // "user" | "default"
+	Slug            string        `bson:"slug,omitempty" json:"slug,omitempty"`
 	CreatedAt       time.Time     `bson:"created_at" json:"created_at"`
 	UpdatedAt       time.Time     `bson:"updated_at" json:"updated_at"`
 }
 
 type WorkflowStepRef struct {
-	StepID     string            `bson:"step_id" json:"step_id"`
+	StepID     string            `bson:"step_id,omitempty" json:"step_id,omitempty"`
+	Inline     *WorkflowStep     `bson:"inline,omitempty" json:"inline,omitempty"`
 	Order      int               `bson:"order" json:"order"`
 	OnFailure  string            `bson:"on_failure" json:"on_failure"` // "stop" | "continue" | "retry"
 	MaxRetries int               `bson:"max_retries" json:"max_retries"`
@@ -42,6 +46,7 @@ type StepOverride struct {
 
 type Workflow struct {
 	ID              bson.ObjectID     `bson:"_id,omitempty" json:"-"`
+	OrgID           string            `bson:"org_id" json:"org_id"`
 	WorkflowID      string            `bson:"workflow_id" json:"workflow_id"`
 	Name            string            `bson:"name" json:"name"`
 	TargetServerIDs []string          `bson:"target_server_ids" json:"target_server_ids"`
@@ -52,10 +57,10 @@ type Workflow struct {
 
 // ResolvedStep is a step frozen into a run snapshot (library step + overrides applied).
 type ResolvedStep struct {
-	Order       int      `bson:"order" json:"order"`
-	Name        string   `bson:"name" json:"name"`
-	Interpreter string   `bson:"interpreter" json:"interpreter"`
-	Script      string   `bson:"script" json:"script"`
+	Order       int               `bson:"order" json:"order"`
+	Name        string            `bson:"name" json:"name"`
+	Interpreter string            `bson:"interpreter" json:"interpreter"`
+	Script      string            `bson:"script" json:"script"`
 	SecretRefs  []string          `bson:"secret_refs" json:"secret_refs"`
 	OnFailure   string            `bson:"on_failure" json:"on_failure"`
 	MaxRetries  int               `bson:"max_retries" json:"max_retries"`
@@ -86,6 +91,7 @@ type ServerRun struct {
 
 type WorkflowRun struct {
 	ID          bson.ObjectID  `bson:"_id,omitempty" json:"-"`
+	OrgID       string         `bson:"org_id" json:"org_id"`
 	RunID       string         `bson:"run_id" json:"run_id"`
 	WorkflowID  string         `bson:"workflow_id" json:"workflow_id"`
 	Name        string         `bson:"name" json:"name"`

server/internal/monitorsched/scheduler.go

@@ -0,0 +1,107 @@
+// Package monitorsched runs server-side monitors on their configured interval
+// and funnels results through services.IngestResult. Agent-run monitors
+// (runner != "server") are excluded — those execute on the agent.
+package monitorsched
+
+import (
+	"context"
+	"log"
+	"sync"
+	"time"
+
+	"github.com/mrhid6/vantage/server/internal/checker"
+	"github.com/mrhid6/vantage/server/internal/models"
+	"github.com/mrhid6/vantage/server/internal/services"
+)
+
+// reloadInterval controls how often the scheduler re-reads monitor definitions
+// so CRUD changes (new/removed/edited monitors) take effect.
+const reloadInterval = 30 * time.Second
+
+type runner struct {
+	monitorID   string
+	intervalSec int
+	cancel      context.CancelFunc
+}
+
+// Start launches the scheduler loop. It returns immediately; the loop runs until
+// ctx is cancelled.
+func Start(ctx context.Context) {
+	go loop(ctx)
+}
+
+func loop(ctx context.Context) {
+	active := map[string]*runner{}
+	var mu sync.Mutex
+
+	sync := func() {
+		monitors, err := services.ListServerScheduledMonitors()
+		if err != nil {
+			log.Printf("monitorsched: list monitors: %v", err)
+			return
+		}
+		want := map[string]models.Monitor{}
+		for _, m := range monitors {
+			want[m.MonitorID] = m
+		}
+
+		mu.Lock()
+		defer mu.Unlock()
+		// Stop runners for monitors that vanished or changed interval.
+		for id, r := range active {
+			m, ok := want[id]
+			if !ok || m.IntervalSec != r.intervalSec {
+				r.cancel()
+				delete(active, id)
+			}
+		}
+		// Start runners for new/changed monitors.
+		for id, m := range want {
+			if _, ok := active[id]; ok {
+				continue
+			}
+			rctx, cancel := context.WithCancel(ctx)
+			active[id] = &runner{monitorID: id, intervalSec: m.IntervalSec, cancel: cancel}
+			go runMonitor(rctx, m)
+		}
+	}
+
+	sync()
+	t := time.NewTicker(reloadInterval)
+	defer t.Stop()
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-t.C:
+			sync()
+		}
+	}
+}
+
+func runMonitor(ctx context.Context, m models.Monitor) {
+	interval := time.Duration(m.IntervalSec) * time.Second
+	if interval <= 0 {
+		interval = 60 * time.Second
+	}
+	spec := services.SpecFor(&m)
+
+	run := func() {
+		res := checker.Run(ctx, spec)
+		if err := services.IngestServerScheduledResult(m.MonitorID, res); err != nil {
+			log.Printf("monitorsched: ingest %s: %v", m.MonitorID, err)
+		}
+	}
+
+	run() // check immediately on (re)start
+	t := time.NewTicker(interval)
+	defer t.Stop()
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-t.C:
+			run()
+		}
+	}
+}

server/internal/notify/dispatch.go

@@ -0,0 +1,64 @@
+// Package notify formats and delivers monitor state-change alerts to
+// notification channels. It depends only on models so services can call it
+// without an import cycle.
+package notify
+
+import (
+	"fmt"
+	"time"
+
+	"github.com/mrhid6/vantage/server/internal/models"
+)
+
+// Event describes a monitor state transition worth alerting on.
+type Event struct {
+	MonitorName string
+	Type        string
+	OldStatus   string
+	NewStatus   string
+	Message     string
+	Time        time.Time
+}
+
+// title is a short one-line summary used by the text-based channels.
+func (e Event) title() string {
+	verb := "recovered"
+	if e.NewStatus == models.StatusDown {
+		verb = "is DOWN"
+	}
+	s := fmt.Sprintf("[Vantage] %s (%s) %s", e.MonitorName, e.Type, verb)
+	if e.Message != "" {
+		s += ": " + e.Message
+	}
+	return s
+}
+
+// Dispatch delivers ev to a single channel, formatting per channel type.
+func Dispatch(ch models.NotificationChannel, ev Event) error {
+	switch ch.Type {
+	case models.ChannelWebhook:
+		return dispatchWebhook(ch, ev)
+	case models.ChannelDiscord:
+		return dispatchDiscord(ch, ev)
+	case models.ChannelSlack:
+		return dispatchSlack(ch, ev)
+	case models.ChannelTelegram:
+		return dispatchTelegram(ch, ev)
+	case models.ChannelSMTP:
+		return dispatchSMTP(ch, ev)
+	default:
+		return fmt.Errorf("unknown channel type: %s", ch.Type)
+	}
+}
+
+// Test delivers a synthetic event so users can verify a channel's configuration.
+func Test(ch models.NotificationChannel) error {
+	return Dispatch(ch, Event{
+		MonitorName: "Test monitor",
+		Type:        "http",
+		OldStatus:   models.StatusUp,
+		NewStatus:   models.StatusDown,
+		Message:     "this is a test alert from Vantage",
+		Time:        time.Now(),
+	})
+}

server/internal/notify/http.go

@@ -0,0 +1,71 @@
+package notify
+
+import (
+	"bytes"
+	"encoding/json"
+	"fmt"
+	"net/http"
+	"time"
+
+	"github.com/mrhid6/vantage/server/internal/models"
+)
+
+var httpClient = &http.Client{Timeout: 10 * time.Second}
+
+func postJSON(target string, payload any) error {
+	body, err := json.Marshal(payload)
+	if err != nil {
+		return err
+	}
+	resp, err := httpClient.Post(target, "application/json", bytes.NewReader(body))
+	if err != nil {
+		return err
+	}
+	defer resp.Body.Close()
+	if resp.StatusCode >= 300 {
+		return fmt.Errorf("HTTP %d from %s", resp.StatusCode, target)
+	}
+	return nil
+}
+
+// dispatchWebhook posts the full event as JSON to a user-supplied URL.
+func dispatchWebhook(ch models.NotificationChannel, ev Event) error {
+	target := ch.Config["url"]
+	if target == "" {
+		return fmt.Errorf("webhook: missing url")
+	}
+	return postJSON(target, map[string]any{
+		"monitor":    ev.MonitorName,
+		"type":       ev.Type,
+		"old_status": ev.OldStatus,
+		"new_status": ev.NewStatus,
+		"message":    ev.Message,
+		"time":       ev.Time.Format(time.RFC3339),
+	})
+}
+
+func dispatchDiscord(ch models.NotificationChannel, ev Event) error {
+	target := ch.Config["url"]
+	if target == "" {
+		return fmt.Errorf("discord: missing url")
+	}
+	return postJSON(target, map[string]string{"content": ev.title()})
+}
+
+func dispatchSlack(ch models.NotificationChannel, ev Event) error {
+	target := ch.Config["url"]
+	if target == "" {
+		return fmt.Errorf("slack: missing url")
+	}
+	return postJSON(target, map[string]string{"text": ev.title()})
+}
+
+func dispatchTelegram(ch models.NotificationChannel, ev Event) error {
+	token := ch.Config["token"]
+	chatID := ch.Config["chat_id"]
+	if token == "" || chatID == "" {
+		return fmt.Errorf("telegram: missing token or chat_id")
+	}
+	api := fmt.Sprintf("https://api.telegram.org/bot%s/sendMessage", token)
+	return postJSON(api, map[string]string{"chat_id": chatID, "text": ev.title()})
+}

server/internal/notify/smtp.go

@@ -0,0 +1,98 @@
+package notify
+
+import (
+	"crypto/tls"
+	"fmt"
+	"net"
+	"net/smtp"
+	"strings"
+	"time"
+
+	"github.com/mrhid6/vantage/server/internal/models"
+)
+
+const smtpTimeout = 15 * time.Second
+
+// dispatchSMTP sends the alert as a plain-text email. Config keys: host, port,
+// username, password, from, to. Auth is skipped when username is empty. Port 465
+// uses implicit TLS; other ports use STARTTLS when the server advertises it.
+//
+// It dials with a timeout and sets a connection deadline so an unreachable or
+// misconfigured SMTP host fails fast instead of hanging the request until the OS
+// TCP timeout (which resets the upstream proxy connection).
+func dispatchSMTP(ch models.NotificationChannel, ev Event) error {
+	host := ch.Config["host"]
+	port := ch.Config["port"]
+	from := ch.Config["from"]
+	to := ch.Config["to"]
+	if host == "" || port == "" || from == "" || to == "" {
+		return fmt.Errorf("smtp: missing host/port/from/to")
+	}
+
+	addr := net.JoinHostPort(host, port)
+	conn, err := net.DialTimeout("tcp", addr, smtpTimeout)
+	if err != nil {
+		return fmt.Errorf("smtp: dial %s: %w", addr, err)
+	}
+	_ = conn.SetDeadline(time.Now().Add(smtpTimeout))
+
+	// Implicit TLS on 465; otherwise start plain and upgrade via STARTTLS.
+	if port == "465" {
+		conn = tls.Client(conn, &tls.Config{ServerName: host})
+	}
+
+	c, err := smtp.NewClient(conn, host)
+	if err != nil {
+		conn.Close()
+		return fmt.Errorf("smtp: client: %w", err)
+	}
+	defer c.Close()
+
+	if port != "465" {
+		if ok, _ := c.Extension("STARTTLS"); ok {
+			if err := c.StartTLS(&tls.Config{ServerName: host}); err != nil {
+				return fmt.Errorf("smtp: starttls: %w", err)
+			}
+		}
+	}
+
+	if user := ch.Config["username"]; user != "" {
+		if err := c.Auth(smtp.PlainAuth("", user, ch.Config["password"], host)); err != nil {
+			return fmt.Errorf("smtp: auth: %w", err)
+		}
+	}
+
+	recipients := strings.Split(to, ",")
+	for i := range recipients {
+		recipients[i] = strings.TrimSpace(recipients[i])
+	}
+
+	if err := c.Mail(from); err != nil {
+		return fmt.Errorf("smtp: mail from: %w", err)
+	}
+	for _, rcpt := range recipients {
+		if rcpt == "" {
+			continue
+		}
+		if err := c.Rcpt(rcpt); err != nil {
+			return fmt.Errorf("smtp: rcpt %s: %w", rcpt, err)
+		}
+	}
+
+	msg, err := buildMIME(from, to, ev.title(), textEmail(ev), htmlEmail(ev))
+	if err != nil {
+		return fmt.Errorf("smtp: build message: %w", err)
+	}
+
+	w, err := c.Data()
+	if err != nil {
+		return fmt.Errorf("smtp: data: %w", err)
+	}
+	if _, err := w.Write(msg); err != nil {
+		return fmt.Errorf("smtp: write: %w", err)
+	}
+	if err := w.Close(); err != nil {
+		return fmt.Errorf("smtp: close data: %w", err)
+	}
+	return c.Quit()
+}

server/internal/notify/template.go

@@ -0,0 +1,167 @@
+package notify
+
+import (
+	"fmt"
+	"html"
+	"mime/multipart"
+	"net/textproto"
+	"strings"
+
+	"github.com/mrhid6/vantage/server/internal/models"
+)
+
+// App theme colors (mirrors web/tailwind.config.ts).
+const (
+	colBg        = "#0f1117"
+	colSurface   = "#1a1d27"
+	colSurface2  = "#232635"
+	colBorder    = "#2e3147"
+	colText      = "#e8eaf0"
+	colTextMuted = "#9095a8"
+	colAccent    = "#6366f1"
+	colSuccess   = "#22c55e"
+	colDanger    = "#ef4444"
+)
+
+// statusColor returns the accent color for a monitor status.
+func statusColor(status string) string {
+	switch status {
+	case models.StatusUp:
+		return colSuccess
+	case models.StatusDown:
+		return colDanger
+	default:
+		return colTextMuted
+	}
+}
+
+// buildMIME assembles a multipart/alternative message (plain + HTML) with the
+// standard email headers, ready to hand to the SMTP DATA command.
+func buildMIME(from, to, subject, text, htmlBody string) ([]byte, error) {
+	var buf strings.Builder
+	w := multipart.NewWriter(&buf)
+
+	var head strings.Builder
+	head.WriteString("From: " + from + "\r\n")
+	head.WriteString("To: " + to + "\r\n")
+	head.WriteString("Subject: " + subject + "\r\n")
+	head.WriteString("MIME-Version: 1.0\r\n")
+	head.WriteString(fmt.Sprintf("Content-Type: multipart/alternative; boundary=%s\r\n\r\n", w.Boundary()))
+
+	textPart, err := w.CreatePart(textproto.MIMEHeader{"Content-Type": {"text/plain; charset=UTF-8"}})
+	if err != nil {
+		return nil, err
+	}
+	textPart.Write([]byte(text))
+
+	htmlPart, err := w.CreatePart(textproto.MIMEHeader{"Content-Type": {"text/html; charset=UTF-8"}})
+	if err != nil {
+		return nil, err
+	}
+	htmlPart.Write([]byte(htmlBody))
+
+	if err := w.Close(); err != nil {
+		return nil, err
+	}
+	return []byte(head.String() + buf.String()), nil
+}
+
+// htmlEmail renders the alert as a dark-themed HTML email matching the app.
+func htmlEmail(ev Event) string {
+	accent := statusColor(ev.NewStatus)
+	label := "Recovered"
+	if ev.NewStatus == models.StatusDown {
+		label = "Down"
+	}
+
+	esc := html.EscapeString
+	row := func(k, v string) string {
+		if v == "" {
+			v = "—"
+		}
+		return fmt.Sprintf(
+			`<tr>`+
+				`<td style="padding:8px 0;color:%s;font-size:13px;width:120px;">%s</td>`+
+				`<td style="padding:8px 0;color:%s;font-size:13px;font-weight:500;">%s</td>`+
+				`</tr>`,
+			colTextMuted, k, colText, esc(v))
+	}
+
+	message := ""
+	if ev.Message != "" {
+		message = fmt.Sprintf(
+			`<p style="margin:0 0 20px;padding:12px 14px;background:%s;border:1px solid %s;border-radius:8px;color:%s;font-size:13px;">%s</p>`,
+			colSurface2, colBorder, colText, esc(ev.Message))
+	}
+
+	transition := esc(ev.OldStatus) + " → " + esc(ev.NewStatus)
+
+	return fmt.Sprintf(`<!DOCTYPE html>
+<html>
+<body style="margin:0;padding:0;background:%s;">
+  <table role="presentation" width="100%%" cellpadding="0" cellspacing="0" style="background:%s;padding:32px 0;">
+    <tr>
+      <td align="center">
+        <table role="presentation" width="480" cellpadding="0" cellspacing="0" style="max-width:480px;background:%s;border:1px solid %s;border-radius:12px;overflow:hidden;font-family:-apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,Helvetica,Arial,sans-serif;">
+          <tr><td style="height:4px;background:%s;"></td></tr>
+          <tr>
+            <td style="padding:28px 28px 20px;">
+              <table role="presentation" cellpadding="0" cellspacing="0" style="margin-bottom:18px;">
+                <tr>
+                  <td style="font-size:18px;font-weight:700;color:%s;">Vantage</td>
+                </tr>
+              </table>
+              <span style="display:inline-block;padding:4px 12px;border-radius:9999px;background:%s22;color:%s;font-size:12px;font-weight:600;letter-spacing:.02em;">%s</span>
+              <h1 style="margin:14px 0 6px;font-size:20px;font-weight:700;color:%s;">%s</h1>
+              <p style="margin:0 0 20px;color:%s;font-size:13px;">%s check</p>
+              %s
+              <table role="presentation" width="100%%" cellpadding="0" cellspacing="0" style="border-top:1px solid %s;">
+                %s
+                %s
+                %s
+              </table>
+            </td>
+          </tr>
+          <tr>
+            <td style="padding:16px 28px;border-top:1px solid %s;">
+              <p style="margin:0;color:%s;font-size:12px;">Sent by Vantage · self-hosted service monitoring</p>
+            </td>
+          </tr>
+        </table>
+      </td>
+    </tr>
+  </table>
+</body>
+</html>`,
+		colBg,
+		colBg,
+		colSurface, colBorder,
+		accent,
+		colText,
+		accent, accent, label,
+		colText, esc(ev.MonitorName),
+		colTextMuted, esc(ev.Type),
+		message,
+		colBorder,
+		row("Status", transition),
+		row("Type", ev.Type),
+		row("Time", ev.Time.Format("2006-01-02 15:04:05 MST")),
+		colBorder,
+		colTextMuted,
+	)
+}
+
+// textEmail renders the plain-text fallback.
+func textEmail(ev Event) string {
+	return strings.Join([]string{
+		ev.title(),
+		"",
+		"Monitor: " + ev.MonitorName,
+		"Type:    " + ev.Type,
+		"Status:  " + ev.OldStatus + " -> " + ev.NewStatus,
+		"Message: " + ev.Message,
+		"Time:    " + ev.Time.Format("2006-01-02 15:04:05 MST"),
+		"",
+		"Sent by Vantage · self-hosted service monitoring",
+	}, "\r\n")
+}

server/internal/services/audit.go

@@ -11,11 +11,12 @@ import (
 	"go.mongodb.org/mongo-driver/v2/mongo/options"
 )
 
-func LogEvent(eventType, actor, serverID, keyID, details string) {
+func LogEvent(orgID, eventType, actor, serverID, keyID, details string) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
 	event := models.AuditEvent{
+		OrgID:     orgID,
 		EventType: eventType,
 		Actor:     actor,
 		ServerID:  serverID,
@@ -28,7 +29,7 @@ func LogEvent(eventType, actor, serverID, keyID, details string) {
 	}
 }
 
-func ListAuditEvents(limit int64) ([]models.AuditEvent, error) {
+func ListAuditEvents(orgID string, limit int64) ([]models.AuditEvent, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
@@ -36,7 +37,7 @@ func ListAuditEvents(limit int64) ([]models.AuditEvent, error) {
 		SetSort(bson.D{{Key: "created_at", Value: -1}}).
 		SetLimit(limit)
 
-	cursor, err := db.Col("audit_logs").Find(ctx, bson.M{}, opts)
+	cursor, err := db.Col("audit_logs").Find(ctx, bson.M{"org_id": orgID}, opts)
 	if err != nil {
 		return nil, err
 	}

server/internal/services/channels.go

@@ -0,0 +1,116 @@
+package services
+
+import (
+	"errors"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/mrhid6/vantage/server/internal/db"
+	"github.com/mrhid6/vantage/server/internal/models"
+	"github.com/mrhid6/vantage/server/internal/notify"
+	"go.mongodb.org/mongo-driver/v2/bson"
+	"go.mongodb.org/mongo-driver/v2/mongo"
+	"go.mongodb.org/mongo-driver/v2/mongo/options"
+)
+
+func ListChannels(orgID string) ([]models.NotificationChannel, error) {
+	ctx, cancel := monCtx()
+	defer cancel()
+	cur, err := db.Col("notification_channels").Find(ctx, bson.M{"org_id": orgID}, options.Find().SetSort(bson.M{"created_at": 1}))
+	if err != nil {
+		return nil, err
+	}
+	var out []models.NotificationChannel
+	if err := cur.All(ctx, &out); err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
+func GetChannel(orgID, channelID string) (*models.NotificationChannel, error) {
+	ctx, cancel := monCtx()
+	defer cancel()
+	var ch models.NotificationChannel
+	err := db.Col("notification_channels").FindOne(ctx, bson.M{"channel_id": channelID, "org_id": orgID}).Decode(&ch)
+	if errors.Is(err, mongo.ErrNoDocuments) {
+		return nil, nil
+	}
+	if err != nil {
+		return nil, err
+	}
+	return &ch, nil
+}
+
+// GetChannels loads multiple channels by ID within an org, skipping any not found.
+func GetChannels(orgID string, channelIDs []string) ([]models.NotificationChannel, error) {
+	if len(channelIDs) == 0 {
+		return nil, nil
+	}
+	ctx, cancel := monCtx()
+	defer cancel()
+	cur, err := db.Col("notification_channels").Find(ctx, bson.M{"org_id": orgID, "channel_id": bson.M{"$in": channelIDs}})
+	if err != nil {
+		return nil, err
+	}
+	var out []models.NotificationChannel
+	if err := cur.All(ctx, &out); err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
+// validateChannelIDs rejects any channel that does not belong to the org.
+// Channel IDs arrive from the client as data on monitor writes.
+func validateChannelIDs(orgID string, channelIDs []string) error {
+	for _, id := range channelIDs {
+		ch, err := GetChannel(orgID, id)
+		if err != nil {
+			return err
+		}
+		if ch == nil {
+			return errors.New("channel " + id + " not found")
+		}
+	}
+	return nil
+}
+
+func CreateChannel(orgID string, ch *models.NotificationChannel) (*models.NotificationChannel, error) {
+	ctx, cancel := monCtx()
+	defer cancel()
+	ch.OrgID = orgID
+	ch.ChannelID = uuid.NewString()
+	ch.CreatedAt = time.Now()
+	if ch.Config == nil {
+		ch.Config = map[string]string{}
+	}
+	if _, err := db.Col("notification_channels").InsertOne(ctx, ch); err != nil {
+		return nil, err
+	}
+	return ch, nil
+}
+
+func UpdateChannel(orgID, channelID string, upd bson.M) error {
+	ctx, cancel := monCtx()
+	defer cancel()
+	_, err := db.Col("notification_channels").UpdateOne(ctx, bson.M{"channel_id": channelID, "org_id": orgID}, bson.M{"$set": upd})
+	return err
+}
+
+func DeleteChannel(orgID, channelID string) error {
+	ctx, cancel := monCtx()
+	defer cancel()
+	_, err := db.Col("notification_channels").DeleteOne(ctx, bson.M{"channel_id": channelID, "org_id": orgID})
+	return err
+}
+
+// TestChannel sends a synthetic alert to verify configuration.
+func TestChannel(orgID, channelID string) error {
+	ch, err := GetChannel(orgID, channelID)
+	if err != nil {
+		return err
+	}
+	if ch == nil {
+		return errors.New("channel not found")
+	}
+	return notify.Test(*ch)
+}

server/internal/services/console.go

@@ -129,11 +129,12 @@ func BuildGuacParams(srv *models.Server, protocol, sshUser, privateKey, passphra
 	}
 }
 
-func CreateConsoleSession(serverID, protocol, keyID, user, clientIP string) (*models.ConsoleSession, error) {
+func CreateConsoleSession(orgID, serverID, protocol, keyID, user, clientIP string) (*models.ConsoleSession, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
 	s := &models.ConsoleSession{
+		OrgID:     orgID,
 		SessionID: uuid.NewString(),
 		ServerID:  serverID,
 		Protocol:  protocol,
@@ -148,11 +149,11 @@ func CreateConsoleSession(serverID, protocol, keyID, user, clientIP string) (*mo
 	return s, nil
 }
 
-func GetConsoleSession(sessionID string) (*models.ConsoleSession, error) {
+func GetConsoleSession(orgID, sessionID string) (*models.ConsoleSession, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 	var s models.ConsoleSession
-	if err := db.Col("console_sessions").FindOne(ctx, bson.M{"session_id": sessionID}).Decode(&s); err != nil {
+	if err := db.Col("console_sessions").FindOne(ctx, bson.M{"session_id": sessionID, "org_id": orgID}).Decode(&s); err != nil {
 		return nil, err
 	}
 	return &s, nil
@@ -160,7 +161,7 @@ func GetConsoleSession(sessionID string) (*models.ConsoleSession, error) {
 
 // StashConsoleRDPCreds encrypts and stores single-use RDP credentials on the
 // session document. They are consumed (and cleared) when the tunnel opens.
-func StashConsoleRDPCreds(sessionID, username, password string) error {
+func StashConsoleRDPCreds(orgID, sessionID, username, password string) error {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 	u, err := encryptString(username)
@@ -172,7 +173,7 @@ func StashConsoleRDPCreds(sessionID, username, password string) error {
 		return err
 	}
 	_, err = db.Col("console_sessions").UpdateOne(ctx,
-		bson.M{"session_id": sessionID},
+		bson.M{"session_id": sessionID, "org_id": orgID},
 		bson.M{"$set": bson.M{"rdp_user_enc": u, "rdp_pass_enc": p}},
 	)
 	return err
@@ -181,8 +182,8 @@ func StashConsoleRDPCreds(sessionID, username, password string) error {
 // ConsumeConsoleRDPCreds decrypts and returns the stored RDP credentials, then
 // clears them from the session document (single-use). Returns empty strings if
 // none were stored.
-func ConsumeConsoleRDPCreds(sessionID string) (username, password string, err error) {
-	s, err := GetConsoleSession(sessionID)
+func ConsumeConsoleRDPCreds(orgID, sessionID string) (username, password string, err error) {
+	s, err := GetConsoleSession(orgID, sessionID)
 	if err != nil {
 		return "", "", err
 	}
@@ -202,18 +203,18 @@ func ConsumeConsoleRDPCreds(sessionID string) (username, password string, err er
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 	_, _ = db.Col("console_sessions").UpdateOne(ctx,
-		bson.M{"session_id": sessionID},
+		bson.M{"session_id": sessionID, "org_id": orgID},
 		bson.M{"$unset": bson.M{"rdp_user_enc": "", "rdp_pass_enc": ""}},
 	)
 	return username, password, nil
 }
 
 // SetConsoleSSHUser persists the SSH username to use on the session doc.
-func SetConsoleSSHUser(sessionID, username string) error {
+func SetConsoleSSHUser(orgID, sessionID, username string) error {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 	_, err := db.Col("console_sessions").UpdateOne(ctx,
-		bson.M{"session_id": sessionID},
+		bson.M{"session_id": sessionID, "org_id": orgID},
 		bson.M{"$set": bson.M{"ssh_username": username}})
 	return err
 }
@@ -221,12 +222,12 @@ func SetConsoleSSHUser(sessionID, username string) error {
 // ConsumeSessionToken atomically marks a session's one-time token as spent.
 // It returns an error if the token was already consumed (replay) or the session
 // does not exist, so the tunnel can be opened at most once per issued token.
-func ConsumeSessionToken(sessionID string) error {
+func ConsumeSessionToken(orgID, sessionID string) error {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 	now := time.Now()
 	res, err := db.Col("console_sessions").UpdateOne(ctx,
-		bson.M{"session_id": sessionID, "token_consumed_at": nil},
+		bson.M{"session_id": sessionID, "org_id": orgID, "token_consumed_at": nil},
 		bson.M{"$set": bson.M{"token_consumed_at": now}},
 	)
 	if err != nil {
@@ -238,12 +239,12 @@ func ConsumeSessionToken(sessionID string) error {
 	return nil
 }
 
-func EndConsoleSession(sessionID string) error {
+func EndConsoleSession(orgID, sessionID string) error {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 	now := time.Now()
 	_, err := db.Col("console_sessions").UpdateOne(ctx,
-		bson.M{"session_id": sessionID, "ended_at": nil},
+		bson.M{"session_id": sessionID, "org_id": orgID, "ended_at": nil},
 		bson.M{"$set": bson.M{"ended_at": now}},
 	)
 	return err

server/internal/services/console_test.go

@@ -1,111 +0,0 @@
-package services
-
-import (
-	"testing"
-	"time"
-
-	"github.com/mrhid6/vantage/server/internal/models"
-)
-
-func TestSessionTokenRoundTrip(t *testing.T) {
-	t.Setenv("KEY_ENCRYPTION_KEY", "00112233445566778899aabbccddeeff00112233445566778899aabbccddeeff")
-
-	tok, err := SignSessionToken("sess-123", time.Minute)
-	if err != nil {
-		t.Fatalf("sign: %v", err)
-	}
-	got, err := VerifySessionToken(tok)
-	if err != nil {
-		t.Fatalf("verify: %v", err)
-	}
-	if got != "sess-123" {
-		t.Fatalf("got %q want sess-123", got)
-	}
-}
-
-func TestSessionTokenExpired(t *testing.T) {
-	t.Setenv("KEY_ENCRYPTION_KEY", "00112233445566778899aabbccddeeff00112233445566778899aabbccddeeff")
-
-	tok, err := SignSessionToken("sess-123", -time.Second)
-	if err != nil {
-		t.Fatalf("sign: %v", err)
-	}
-	if _, err := VerifySessionToken(tok); err == nil {
-		t.Fatalf("expected expiry error, got nil")
-	}
-}
-
-func TestSessionTokenTampered(t *testing.T) {
-	t.Setenv("KEY_ENCRYPTION_KEY", "00112233445566778899aabbccddeeff00112233445566778899aabbccddeeff")
-
-	tok, _ := SignSessionToken("sess-123", time.Minute)
-	if _, err := VerifySessionToken(tok + "x"); err == nil {
-		t.Fatalf("expected signature error, got nil")
-	}
-}
-
-func TestBuildGuacParamsSSH(t *testing.T) {
-	srv := &models.Server{IPAddress: "10.0.0.5", SSHPort: 22}
-	p, err := BuildGuacParams(srv, "ssh", "", "PRIVATE-KEY-DATA", "", "", "")
-	if err != nil {
-		t.Fatalf("err: %v", err)
-	}
-	if p.Protocol != "ssh" {
-		t.Fatalf("protocol %q", p.Protocol)
-	}
-	if p.Params["hostname"] != "10.0.0.5" || p.Params["port"] != "22" {
-		t.Fatalf("bad host/port: %+v", p.Params)
-	}
-	if p.Params["private-key"] != "PRIVATE-KEY-DATA" {
-		t.Fatalf("missing private-key")
-	}
-	if p.Params["username"] != "root" {
-		t.Fatalf("expected default username root, got %q", p.Params["username"])
-	}
-}
-
-func TestBuildGuacParamsRDP(t *testing.T) {
-	srv := &models.Server{IPAddress: "10.0.0.9", RDPPort: 3389}
-	p, err := BuildGuacParams(srv, "rdp", "", "", "", "administrator", "s3cret")
-	if err != nil {
-		t.Fatalf("err: %v", err)
-	}
-	if p.Params["port"] != "3389" || p.Params["username"] != "administrator" || p.Params["password"] != "s3cret" {
-		t.Fatalf("bad rdp params: %+v", p.Params)
-	}
-	if p.Params["ignore-cert"] != "true" {
-		t.Fatalf("expected ignore-cert=true")
-	}
-}
-
-func TestBuildGuacParamsUnknownProtocol(t *testing.T) {
-	srv := &models.Server{IPAddress: "10.0.0.9"}
-	if _, err := BuildGuacParams(srv, "telnet", "", "", "", "", ""); err == nil {
-		t.Fatalf("expected error for unknown protocol")
-	}
-}
-
-func TestBuildGuacParamsSSHPassphrase(t *testing.T) {
-	srv := &models.Server{IPAddress: "10.0.0.5", SSHPort: 22}
-	p, err := BuildGuacParams(srv, "ssh", "deploy", "PK", "s3cret-phrase", "", "")
-	if err != nil {
-		t.Fatalf("err: %v", err)
-	}
-	if p.Params["username"] != "deploy" {
-		t.Fatalf("username %q", p.Params["username"])
-	}
-	if p.Params["passphrase"] != "s3cret-phrase" {
-		t.Fatalf("missing passphrase: %+v", p.Params)
-	}
-}
-
-func TestBuildGuacParamsVNC(t *testing.T) {
-	srv := &models.Server{IPAddress: "10.0.0.7"}
-	p, err := BuildGuacParams(srv, "vnc", "", "", "", "", "vncpass")
-	if err != nil {
-		t.Fatalf("err: %v", err)
-	}
-	if p.Protocol != "vnc" || p.Params["hostname"] != "10.0.0.7" || p.Params["port"] != "5900" || p.Params["password"] != "vncpass" {
-		t.Fatalf("bad vnc params: %+v", p.Params)
-	}
-}

server/internal/services/defaults.go

@@ -0,0 +1,95 @@
+package services
+
+import (
+	"os"
+	"path/filepath"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/mrhid6/vantage/server/internal/db"
+	"github.com/mrhid6/vantage/server/internal/models"
+	"go.mongodb.org/mongo-driver/v2/bson"
+	"go.mongodb.org/mongo-driver/v2/mongo/options"
+)
+
+// DefaultStepsDir returns the directory holding default step JSON files.
+func DefaultStepsDir() string {
+	dir := os.Getenv("VANTAGE_DEFAULT_STEPS_DIR")
+	if dir == "" {
+		dir = filepath.Join("data", "default-steps")
+	}
+	_ = os.MkdirAll(dir, 0700)
+	return dir
+}
+
+// readDefaultStepFiles parses every *.json in the defaults dir into
+// source=default library steps (with slug set). Non-json and invalid files are
+// skipped silently; a slug is derived from the step name.
+func readDefaultStepFiles() ([]models.WorkflowStep, error) {
+	matches, err := filepath.Glob(filepath.Join(DefaultStepsDir(), "*.json"))
+	if err != nil {
+		return nil, err
+	}
+	out := []models.WorkflowStep{}
+	for _, path := range matches {
+		b, err := os.ReadFile(path)
+		if err != nil {
+			continue
+		}
+		s, err := ParseStepDoc(b)
+		if err != nil {
+			continue
+		}
+		s.Source = "default"
+		s.Slug = Slugify(s.Name)
+		if s.Slug == "" {
+			continue
+		}
+		out = append(out, s)
+	}
+	return out, nil
+}
+
+// SeedDefaultSteps upserts default steps from disk keyed on {slug, source}.
+// Re-sync overwrites default-step content; user steps are never touched.
+func SeedDefaultSteps(orgID string) (created, updated int, err error) {
+	steps, err := readDefaultStepFiles()
+	if err != nil {
+		return 0, 0, err
+	}
+	ctx, cancel := wfCtx()
+	defer cancel()
+	col := db.Col("workflow_steps")
+	for _, s := range steps {
+		filter := bson.M{"org_id": orgID, "slug": s.Slug, "source": "default"}
+		set := bson.M{
+			"name":             s.Name,
+			"description":      s.Description,
+			"interpreter":      s.Interpreter,
+			"script":           s.Script,
+			"declared_outputs": s.DeclaredOutputs,
+			"declared_inputs":  s.DeclaredInputs,
+			"secret_refs":      s.SecretRefs,
+			"updated_at":       time.Now(),
+		}
+		res, uerr := col.UpdateOne(ctx, filter, bson.M{
+			"$set": set,
+			"$setOnInsert": bson.M{
+				"org_id":     orgID,
+				"step_id":    uuid.New().String(),
+				"slug":       s.Slug,
+				"source":     "default",
+				"created_at": time.Now(),
+			},
+		}, options.UpdateOne().SetUpsert(true))
+		if uerr != nil {
+			return created, updated, uerr
+		}
+		if res.UpsertedCount > 0 {
+			created++
+		} else if res.ModifiedCount > 0 {
+			updated++
+		}
+	}
+	return created, updated, nil
+}

server/internal/services/inventory.go

@@ -0,0 +1,52 @@
+package services
+
+import (
+	"context"
+	"time"
+
+	"github.com/mrhid6/vantage/server/internal/db"
+	"github.com/mrhid6/vantage/server/internal/grpc/pb"
+	"go.mongodb.org/mongo-driver/v2/bson"
+)
+
+// StoreInventory upserts the latest inventory snapshot onto the server document.
+// Metrics fields update every call; static fields only when r.IncludeStatic.
+func StoreInventory(serverID string, r *pb.InventoryReport) error {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+
+	now := time.Now()
+	set := bson.M{"inventory.metrics_at": now}
+	if r.CPU != nil {
+		set["inventory.cpu.usage_pct"] = r.CPU.UsagePct
+		set["inventory.cpu.load1"] = r.CPU.Load1
+	}
+	if r.Memory != nil {
+		set["inventory.memory.used_bytes"] = r.Memory.UsedBytes
+	}
+	set["inventory.swap_used_bytes"] = r.SwapUsed
+
+	if r.IncludeStatic {
+		set["inventory.static_at"] = now
+		set["inventory.swap_total_bytes"] = r.SwapTotal
+		set["inventory.kernel"] = r.Kernel
+		if r.CPU != nil {
+			set["inventory.cpu.model"] = r.CPU.Model
+			set["inventory.cpu.cores"] = r.CPU.Cores
+		}
+		if r.Memory != nil {
+			set["inventory.memory.total_bytes"] = r.Memory.TotalBytes
+		}
+		parts := make([]bson.M, 0, len(r.Partitions))
+		for _, p := range r.Partitions {
+			parts = append(parts, bson.M{
+				"device": p.Device, "mountpoint": p.Mountpoint, "fstype": p.Fstype,
+				"total_bytes": p.TotalBytes, "used_bytes": p.UsedBytes,
+			})
+		}
+		set["inventory.partitions"] = parts
+	}
+
+	_, err := db.Col("servers").UpdateOne(ctx, bson.M{"server_id": serverID}, bson.M{"$set": set})
+	return err
+}

server/internal/services/keys.go

@@ -36,8 +36,9 @@ func setKeyMeta(k *models.Key) {
 	k.HasPassphrase = k.PassphraseEncrypted != ""
 }
 
-func CreateKey(label, publicKey, source, generatedByServerID, privateKey, passphrase string) (*models.Key, error) {
+func CreateKey(orgID, label, publicKey, source, generatedByServerID, privateKey, passphrase string) (*models.Key, error) {
 	key := &models.Key{
+		OrgID:               orgID,
 		KeyID:               uuid.NewString(),
 		Label:               label,
 		PublicKey:           publicKey,
@@ -71,12 +72,12 @@ func CreateKey(label, publicKey, source, generatedByServerID, privateKey, passph
 	return key, nil
 }
 
-func GetKey(keyID string) (*models.Key, error) {
+func GetKey(orgID, keyID string) (*models.Key, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
 	var key models.Key
-	err := db.Col("keys").FindOne(ctx, bson.M{"key_id": keyID}).Decode(&key)
+	err := db.Col("keys").FindOne(ctx, bson.M{"key_id": keyID, "org_id": orgID}).Decode(&key)
 	if err != nil {
 		return nil, err
 	}
@@ -84,12 +85,12 @@ func GetKey(keyID string) (*models.Key, error) {
 	return &key, nil
 }
 
-func GetPrivateKey(keyID string) (string, error) {
+func GetPrivateKey(orgID, keyID string) (string, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
 	var key models.Key
-	if err := db.Col("keys").FindOne(ctx, bson.M{"key_id": keyID}).Decode(&key); err != nil {
+	if err := db.Col("keys").FindOne(ctx, bson.M{"key_id": keyID, "org_id": orgID}).Decode(&key); err != nil {
 		return "", err
 	}
 	if key.PrivateKeyEncrypted == "" {
@@ -99,7 +100,8 @@ func GetPrivateKey(keyID string) (string, error) {
 }
 
 // GetPassphrase returns the decrypted passphrase for a key, or an empty string
-// if the key has none stored.
+// if the key has none stored. Agent-path (keyed by unique key_id from an
+// assignment lookup) — no org filter.
 func GetPassphrase(keyID string) (string, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
@@ -119,11 +121,11 @@ type KeyWithCount struct {
 	AssignedCount int `bson:"-" json:"assigned_count"`
 }
 
-func ListKeys() ([]KeyWithCount, error) {
+func ListKeys(orgID string) ([]KeyWithCount, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
-	cursor, err := db.Col("keys").Find(ctx, bson.M{})
+	cursor, err := db.Col("keys").Find(ctx, bson.M{"org_id": orgID})
 	if err != nil {
 		return nil, err
 	}
@@ -138,6 +140,7 @@ func ListKeys() ([]KeyWithCount, error) {
 	for _, k := range keys {
 		setKeyMeta(&k)
 		count, _ := db.Col("assignments").CountDocuments(ctx, bson.M{
+			"org_id":     orgID,
 			"key_id":     k.KeyID,
 			"revoked_at": nil,
 		})
@@ -146,19 +149,19 @@ func ListKeys() ([]KeyWithCount, error) {
 	return result, nil
 }
 
-func DeleteKey(keyID string) error {
+func DeleteKey(orgID, keyID string) error {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
 	var key models.Key
-	if err := db.Col("keys").FindOne(ctx, bson.M{"key_id": keyID}).Decode(&key); err != nil {
+	if err := db.Col("keys").FindOne(ctx, bson.M{"key_id": keyID, "org_id": orgID}).Decode(&key); err != nil {
 		return err
 	}
 
-	if _, err := db.Col("keys").DeleteOne(ctx, bson.M{"key_id": keyID}); err != nil {
+	if _, err := db.Col("keys").DeleteOne(ctx, bson.M{"key_id": keyID, "org_id": orgID}); err != nil {
 		return err
 	}
-	if _, err := db.Col("assignments").DeleteMany(ctx, bson.M{"key_id": keyID}); err != nil {
+	if _, err := db.Col("assignments").DeleteMany(ctx, bson.M{"key_id": keyID, "org_id": orgID}); err != nil {
 		return err
 	}
 
@@ -168,13 +171,23 @@ func DeleteKey(keyID string) error {
 	return nil
 }
 
-func AssignKey(keyID, serverID string) (*models.Assignment, error) {
+func AssignKey(orgID, keyID, serverID string) (*models.Assignment, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
+	// Both sides must belong to the caller's org — the IDs arrive from the
+	// client as data and are consumed by unscoped agent-path queries later.
+	if _, err := GetKey(orgID, keyID); err != nil {
+		return nil, fmt.Errorf("key not found")
+	}
+	if _, err := GetServer(orgID, serverID); err != nil {
+		return nil, fmt.Errorf("server not found")
+	}
+
 	// Check if already assigned and active
 	var existing models.Assignment
 	err := db.Col("assignments").FindOne(ctx, bson.M{
+		"org_id":     orgID,
 		"key_id":     keyID,
 		"server_id":  serverID,
 		"revoked_at": nil,
@@ -184,6 +197,7 @@ func AssignKey(keyID, serverID string) (*models.Assignment, error) {
 	}
 
 	a := &models.Assignment{
+		OrgID:      orgID,
 		KeyID:      keyID,
 		ServerID:   serverID,
 		AssignedAt: time.Now(),
@@ -195,23 +209,23 @@ func AssignKey(keyID, serverID string) (*models.Assignment, error) {
 	return a, nil
 }
 
-func RevokeAssignment(keyID, serverID string) error {
+func RevokeAssignment(orgID, keyID, serverID string) error {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
 	now := time.Now()
 	_, err := db.Col("assignments").UpdateOne(ctx,
-		bson.M{"key_id": keyID, "server_id": serverID, "revoked_at": nil},
+		bson.M{"org_id": orgID, "key_id": keyID, "server_id": serverID, "revoked_at": nil},
 		bson.M{"$set": bson.M{"revoked_at": now}},
 	)
 	return err
 }
 
-func GetAssignmentsForKey(keyID string) ([]models.Assignment, error) {
+func GetAssignmentsForKey(orgID, keyID string) ([]models.Assignment, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
-	cursor, err := db.Col("assignments").Find(ctx, bson.M{"key_id": keyID, "revoked_at": nil})
+	cursor, err := db.Col("assignments").Find(ctx, bson.M{"org_id": orgID, "key_id": keyID, "revoked_at": nil})
 	if err != nil {
 		return nil, err
 	}
@@ -229,11 +243,11 @@ type AssignmentWithServer struct {
 	Server            *models.Server `json:"server,omitempty"`
 }
 
-func GetAssignmentsWithServers(keyID string) ([]AssignmentWithServer, error) {
+func GetAssignmentsWithServers(orgID, keyID string) ([]AssignmentWithServer, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
-	cursor, err := db.Col("assignments").Find(ctx, bson.M{"key_id": keyID})
+	cursor, err := db.Col("assignments").Find(ctx, bson.M{"org_id": orgID, "key_id": keyID})
 	if err != nil {
 		return nil, err
 	}
@@ -248,7 +262,7 @@ func GetAssignmentsWithServers(keyID string) ([]AssignmentWithServer, error) {
 	for _, a := range assignments {
 		item := AssignmentWithServer{Assignment: a}
 		var srv models.Server
-		if err := db.Col("servers").FindOne(ctx, bson.M{"server_id": a.ServerID}).Decode(&srv); err == nil {
+		if err := db.Col("servers").FindOne(ctx, bson.M{"server_id": a.ServerID, "org_id": orgID}).Decode(&srv); err == nil {
 			item.Server = &srv
 		}
 		result = append(result, item)
@@ -261,11 +275,11 @@ type AssignmentWithKey struct {
 	Key               *models.Key `json:"key,omitempty"`
 }
 
-func GetAssignmentsWithKeysForServer(serverID string) ([]AssignmentWithKey, error) {
+func GetAssignmentsWithKeysForServer(orgID, serverID string) ([]AssignmentWithKey, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
-	cursor, err := db.Col("assignments").Find(ctx, bson.M{"server_id": serverID})
+	cursor, err := db.Col("assignments").Find(ctx, bson.M{"org_id": orgID, "server_id": serverID})
 	if err != nil {
 		return nil, err
 	}
@@ -279,7 +293,7 @@ func GetAssignmentsWithKeysForServer(serverID string) ([]AssignmentWithKey, erro
 	result := make([]AssignmentWithKey, 0, len(assignments))
 	for _, a := range assignments {
 		var key models.Key
-		if err := db.Col("keys").FindOne(ctx, bson.M{"key_id": a.KeyID}).Decode(&key); err != nil {
+		if err := db.Col("keys").FindOne(ctx, bson.M{"key_id": a.KeyID, "org_id": orgID}).Decode(&key); err != nil {
 			continue
 		}
 		setKeyMeta(&key)

server/internal/services/migrate.go

@@ -0,0 +1,269 @@
+package services
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/mrhid6/vantage/server/internal/db"
+	"github.com/mrhid6/vantage/server/internal/models"
+	"go.mongodb.org/mongo-driver/v2/bson"
+	"go.mongodb.org/mongo-driver/v2/mongo"
+	"go.mongodb.org/mongo-driver/v2/mongo/options"
+)
+
+var scopedCollections = []string{
+	"servers", "keys", "assignments", "secrets",
+	"workflows", "workflow_steps", "workflow_runs",
+	"audit_logs", "monitors", "notification_channels",
+	"console_sessions", "incidents", "monitor_rollups",
+}
+
+// EnsureAuthIndexes creates unique indexes for the new auth collections.
+func EnsureAuthIndexes() error {
+	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+	defer cancel()
+
+	if _, err := db.Col("users").Indexes().CreateOne(ctx, mongo.IndexModel{
+		Keys:    bson.D{{Key: "email", Value: 1}},
+		Options: options.Index().SetUnique(true),
+	}); err != nil {
+		return err
+	}
+	if _, err := db.Col("orgs").Indexes().CreateOne(ctx, mongo.IndexModel{
+		Keys:    bson.D{{Key: "slug", Value: 1}},
+		Options: options.Index().SetUnique(true),
+	}); err != nil {
+		return err
+	}
+	if _, err := db.Col("org_oidc").Indexes().CreateOne(ctx, mongo.IndexModel{
+		Keys:    bson.D{{Key: "org_id", Value: 1}},
+		Options: options.Index().SetUnique(true),
+	}); err != nil {
+		return err
+	}
+	return nil
+}
+
+// defaultBackfillOrg resolves the org that org-less legacy documents belong to:
+// the "default" org, created if absent. Shared by 0001 and 0003 so an instance
+// that ran either one converges on the same org.
+func defaultBackfillOrg(ctx context.Context) (*models.Org, error) {
+	var org models.Org
+	err := db.Col("orgs").FindOne(ctx, bson.M{"slug": "default"}).Decode(&org)
+	switch {
+	case err == nil:
+	case errors.Is(err, mongo.ErrNoDocuments):
+		// Only a genuine absence justifies an insert. Treating a timeout or a
+		// decode failure as "absent" would race the fatal unique orgs.slug index
+		// and turn a transient blip into a boot crash.
+		org = models.Org{OrgID: uuid.NewString(), Name: "Default", Slug: "default", CreatedAt: time.Now()}
+		if _, err := db.Col("orgs").InsertOne(ctx, org); err != nil {
+			return nil, err
+		}
+	default:
+		return nil, err
+	}
+	return &org, nil
+}
+
+// RunMigrations backfills a default org onto pre-existing documents. Idempotent
+// via a marker in the migrations collection.
+func RunMigrations() error {
+	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
+	defer cancel()
+
+	const marker = "0001_default_org_backfill"
+	if n, _ := db.Col("migrations").CountDocuments(ctx, bson.M{"_id": marker}); n > 0 {
+		return nil
+	}
+
+	// Only backfill if there is legacy data lacking org_id.
+	needs := false
+	for _, col := range scopedCollections {
+		n, _ := db.Col(col).CountDocuments(ctx, bson.M{"org_id": bson.M{"$exists": false}})
+		if n > 0 {
+			needs = true
+			break
+		}
+	}
+
+	if needs {
+		org, err := defaultBackfillOrg(ctx)
+		if err != nil {
+			return err
+		}
+		for _, col := range scopedCollections {
+			if _, err := db.Col(col).UpdateMany(ctx,
+				bson.M{"org_id": bson.M{"$exists": false}},
+				bson.M{"$set": bson.M{"org_id": org.OrgID}},
+			); err != nil {
+				return err
+			}
+		}
+	}
+
+	_, err := db.Col("migrations").InsertOne(ctx, bson.M{"_id": marker, "applied_at": time.Now()})
+	return err
+}
+
+// MigrateMissedOrgScopes repairs collections that migration 0001 could not
+// reach. 0001 originally listed "audit" and "channels", but the real collections
+// are audit_logs and notification_channels, so on any instance that ran that
+// version those documents were left without org_id — invisible to org-filtered
+// reads, and in the channels' case silently non-firing. The 0001 marker is
+// already written there, so renaming alone does not repair them; this migration
+// converges both the never-migrated and the incorrectly-migrated case.
+//
+// It also stamps console_sessions, incidents and monitor_rollups, which gained
+// an org_id only after 0001 shipped. Those carry an owning monitor/server whose
+// org is authoritative, so they are derived rather than defaulted. Idempotent
+// via a marker in the migrations collection.
+func MigrateMissedOrgScopes() error {
+	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
+	defer cancel()
+
+	const marker = "0003_missed_org_scopes"
+	if n, _ := db.Col("migrations").CountDocuments(ctx, bson.M{"_id": marker}); n > 0 {
+		return nil
+	}
+
+	// Same org resolution 0001 uses, for the collections it meant to cover.
+	missed := []string{"audit_logs", "notification_channels"}
+	needs := false
+	for _, col := range missed {
+		n, _ := db.Col(col).CountDocuments(ctx, bson.M{"org_id": bson.M{"$exists": false}})
+		if n > 0 {
+			needs = true
+			break
+		}
+	}
+	if needs {
+		org, err := defaultBackfillOrg(ctx)
+		if err != nil {
+			return err
+		}
+		for _, col := range missed {
+			if _, err := db.Col(col).UpdateMany(ctx,
+				bson.M{"org_id": bson.M{"$exists": false}},
+				bson.M{"$set": bson.M{"org_id": org.OrgID}},
+			); err != nil {
+				return err
+			}
+		}
+	}
+
+	// Derived from the owning record — defaulting these would hand one org
+	// another org's console history and incident timeline.
+	if err := backfillOrgFromOwner(ctx, "console_sessions", "server_id", "servers", "server_id"); err != nil {
+		return err
+	}
+	if err := backfillOrgFromOwner(ctx, "incidents", "monitor_id", "monitors", "monitor_id"); err != nil {
+		return err
+	}
+	if err := backfillOrgFromOwner(ctx, "monitor_rollups", "monitor_id", "monitors", "monitor_id"); err != nil {
+		return err
+	}
+
+	_, err := db.Col("migrations").InsertOne(ctx, bson.M{"_id": marker, "applied_at": time.Now()})
+	return err
+}
+
+// backfillOrgFromOwner stamps org_id on every doc in col that lacks one, taking
+// the org from the record in ownerCol it points at. Orphans (owner already
+// deleted) are left alone; they are unreachable either way.
+func backfillOrgFromOwner(ctx context.Context, col, localField, ownerCol, ownerField string) error {
+	// Decoded loosely: a single null or non-string value in the collection would
+	// fail a []string decode and abort the migration — and therefore boot — over
+	// one unusable document. Skip what we cannot use instead.
+	var raw []bson.RawValue
+	if err := db.Col(col).Distinct(ctx, localField,
+		bson.M{"org_id": bson.M{"$exists": false}}).Decode(&raw); err != nil {
+		return err
+	}
+	for _, rv := range raw {
+		id, ok := rv.StringValueOK()
+		if !ok || id == "" {
+			continue
+		}
+		var owner struct {
+			OrgID string `bson:"org_id"`
+		}
+		if err := db.Col(ownerCol).FindOne(ctx, bson.M{ownerField: id}).Decode(&owner); err != nil {
+			continue
+		}
+		if owner.OrgID == "" {
+			continue
+		}
+		if _, err := db.Col(col).UpdateMany(ctx,
+			bson.M{localField: id, "org_id": bson.M{"$exists": false}},
+			bson.M{"$set": bson.M{"org_id": owner.OrgID}},
+		); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+// MigrateSettingsOrg stamps the legacy global settings singleton with the
+// default org's ID. Without it an upgrade would orphan the existing SMTP
+// config, alert config, retention setting, and ESO read token. Idempotent via
+// a marker in the migrations collection.
+func MigrateSettingsOrg() error {
+	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
+	defer cancel()
+
+	const marker = "0002_settings_org_backfill"
+	if n, _ := db.Col("migrations").CountDocuments(ctx, bson.M{"_id": marker}); n > 0 {
+		return nil
+	}
+
+	n, _ := db.Col("settings").CountDocuments(ctx, bson.M{"org_id": bson.M{"$exists": false}})
+	if n > 0 {
+		// The org-less settings doc belongs to whichever org already exists —
+		// migration 0001 only creates a "default" org when there was legacy
+		// data to backfill, so keying off that slug would invent a phantom org
+		// and move the real org's config onto it.
+		var org models.Org
+		orgCount, err := db.Col("orgs").CountDocuments(ctx, bson.M{})
+		if err != nil {
+			return err
+		}
+		switch orgCount {
+		case 1:
+			if err := db.Col("orgs").FindOne(ctx, bson.M{}).Decode(&org); err != nil {
+				return err
+			}
+		case 0:
+			org = models.Org{OrgID: uuid.NewString(), Name: "Default", Slug: "default", CreatedAt: time.Now()}
+			if _, err := db.Col("orgs").InsertOne(ctx, org); err != nil {
+				return err
+			}
+		default:
+			// Ambiguous: several orgs but an unstamped settings doc. Guessing
+			// would hand one org another's SMTP config and ESO token. Continuing
+			// is not an option either: the unique settings.org_id index built
+			// straight after this indexes every unstamped doc as null, so two or
+			// more of them collide and boot fails there instead — with a far less
+			// useful message. Stop here, where we can name the remedy.
+			return fmt.Errorf(
+				"settings org migration: %d settings document(s) have no org_id but %d orgs exist; "+
+					"cannot infer the owner. Set org_id manually on each settings document "+
+					"(db.settings.updateOne({_id:<id>},{$set:{org_id:\"<org uuid>\"}})), deleting any "+
+					"duplicates, then restart", n, orgCount)
+		}
+		if org.OrgID != "" {
+			if _, err := db.Col("settings").UpdateMany(ctx,
+				bson.M{"org_id": bson.M{"$exists": false}},
+				bson.M{"$set": bson.M{"org_id": org.OrgID}},
+			); err != nil {
+				return err
+			}
+		}
+	}
+
+	_, err := db.Col("migrations").InsertOne(ctx, bson.M{"_id": marker, "applied_at": time.Now()})
+	return err
+}

server/internal/services/monitors.go

@@ -0,0 +1,407 @@
+package services
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"log"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/mrhid6/vantage/server/internal/checker"
+	"github.com/mrhid6/vantage/server/internal/db"
+	"github.com/mrhid6/vantage/server/internal/models"
+	"github.com/mrhid6/vantage/server/internal/notify"
+	"go.mongodb.org/mongo-driver/v2/bson"
+	"go.mongodb.org/mongo-driver/v2/mongo"
+	"go.mongodb.org/mongo-driver/v2/mongo/options"
+)
+
+func monCtx() (context.Context, context.CancelFunc) {
+	return context.WithTimeout(context.Background(), 5*time.Second)
+}
+
+// SpecFor maps a monitor onto a checker.Spec.
+func SpecFor(m *models.Monitor) checker.Spec {
+	return checker.Spec{
+		Type:           m.Type,
+		URL:            m.Target.URL,
+		Host:           m.Target.Host,
+		Port:           m.Target.Port,
+		Method:         m.Target.Method,
+		ExpectedStatus: m.Target.ExpectedStatus,
+		Keyword:        m.Target.Keyword,
+		TLSWarnDays:    m.Target.TLSWarnDays,
+		Insecure:       m.Target.Insecure,
+		TimeoutSec:     m.IntervalSec,
+	}
+}
+
+func ListMonitors(orgID string) ([]models.Monitor, error) {
+	ctx, cancel := monCtx()
+	defer cancel()
+	cur, err := db.Col("monitors").Find(ctx, bson.M{"org_id": orgID}, options.Find().SetSort(bson.M{"created_at": 1}))
+	if err != nil {
+		return nil, err
+	}
+	var out []models.Monitor
+	if err := cur.All(ctx, &out); err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
+// ListMonitorsForRunner returns enabled monitors whose Runner matches runner,
+// scoped to orgID. Runner is client-supplied at write time, so an agent fetching
+// its own work must scope by the org of its authenticated server record —
+// otherwise another org could point a monitor at that server_id and have it run
+// their checks. An empty orgID is rejected: it would silently widen the query to
+// every org.
+func ListMonitorsForRunner(orgID, runner string) ([]models.Monitor, error) {
+	if orgID == "" {
+		return nil, errors.New("org id required")
+	}
+	return listMonitorsForRunner(orgID, runner)
+}
+
+// ListServerScheduledMonitors returns every enabled server-run monitor across
+// all orgs. This is the in-process scheduler's entry point (mirrors the cross-org
+// MarkOfflineServers sweep) and must never be called from a request-driven path —
+// it performs no org scoping at all.
+func ListServerScheduledMonitors() ([]models.Monitor, error) {
+	return listMonitorsForRunner("", models.RunnerServer)
+}
+
+// listMonitorsForRunner is the shared query. An empty orgID means no org filter
+// and is only reachable via ListServerScheduledMonitors.
+func listMonitorsForRunner(orgID, runner string) ([]models.Monitor, error) {
+	ctx, cancel := monCtx()
+	defer cancel()
+	filter := bson.M{"runner": runner, "enabled": true}
+	if orgID != "" {
+		filter["org_id"] = orgID
+	}
+	cur, err := db.Col("monitors").Find(ctx, filter)
+	if err != nil {
+		return nil, err
+	}
+	var out []models.Monitor
+	if err := cur.All(ctx, &out); err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
+// GetMonitor looks up a monitor scoped to an org (handler/session use).
+func GetMonitor(orgID, monitorID string) (*models.Monitor, error) {
+	ctx, cancel := monCtx()
+	defer cancel()
+	var m models.Monitor
+	err := db.Col("monitors").FindOne(ctx, bson.M{"monitor_id": monitorID, "org_id": orgID}).Decode(&m)
+	if errors.Is(err, mongo.ErrNoDocuments) {
+		return nil, nil
+	}
+	if err != nil {
+		return nil, err
+	}
+	return &m, nil
+}
+
+// getMonitorByID looks up a monitor by its unique monitor_id with no org
+// filter. For agent/scheduler use only (IngestResult), which has no session.
+func getMonitorByID(monitorID string) (*models.Monitor, error) {
+	ctx, cancel := monCtx()
+	defer cancel()
+	var m models.Monitor
+	err := db.Col("monitors").FindOne(ctx, bson.M{"monitor_id": monitorID}).Decode(&m)
+	if errors.Is(err, mongo.ErrNoDocuments) {
+		return nil, nil
+	}
+	if err != nil {
+		return nil, err
+	}
+	return &m, nil
+}
+
+// validateRunner rejects a runner that is neither the reserved server-scheduler
+// value nor a server in the org. The value is client-supplied and is later
+// consumed by an agent's own monitor fetch, so ownership has to be proven at
+// the write boundary.
+func validateRunner(orgID, runner string) error {
+	if runner == "" || runner == models.RunnerServer {
+		return nil
+	}
+	if _, err := GetServer(orgID, runner); err != nil {
+		return fmt.Errorf("runner server %s not found", runner)
+	}
+	return nil
+}
+
+func CreateMonitor(orgID string, m *models.Monitor) (*models.Monitor, error) {
+	ctx, cancel := monCtx()
+	defer cancel()
+	if err := validateChannelIDs(orgID, m.ChannelIDs); err != nil {
+		return nil, err
+	}
+	if err := validateRunner(orgID, m.Runner); err != nil {
+		return nil, err
+	}
+	m.OrgID = orgID
+	m.MonitorID = uuid.NewString()
+	m.CreatedAt = time.Now()
+	if m.IntervalSec <= 0 {
+		m.IntervalSec = 60
+	}
+	if m.Retries <= 0 {
+		m.Retries = 1
+	}
+	if m.Runner == "" {
+		m.Runner = models.RunnerServer
+	}
+	m.State = models.MonitorState{Status: models.StatusPending}
+	if _, err := db.Col("monitors").InsertOne(ctx, m); err != nil {
+		return nil, err
+	}
+	return m, nil
+}
+
+func UpdateMonitor(orgID, monitorID string, upd bson.M) error {
+	ctx, cancel := monCtx()
+	defer cancel()
+	// A present-but-wrong-type value is a hard error: silently skipping the
+	// check would still let the unvalidated value through to the $set.
+	if raw, present := upd["channel_ids"]; present {
+		ids, ok := raw.([]string)
+		if !ok {
+			return fmt.Errorf("channel_ids must be a string array")
+		}
+		if err := validateChannelIDs(orgID, ids); err != nil {
+			return err
+		}
+	}
+	if raw, present := upd["runner"]; present {
+		runner, ok := raw.(string)
+		if !ok {
+			return fmt.Errorf("runner must be a string")
+		}
+		if err := validateRunner(orgID, runner); err != nil {
+			return err
+		}
+		// Match CreateMonitor: an empty runner means the server scheduler.
+		// Storing "" would match no runner at all and silently stop the
+		// monitor being checked.
+		if runner == "" {
+			upd["runner"] = models.RunnerServer
+		}
+	}
+	_, err := db.Col("monitors").UpdateOne(ctx, bson.M{"monitor_id": monitorID, "org_id": orgID}, bson.M{"$set": upd})
+	return err
+}
+
+func DeleteMonitor(orgID, monitorID string) error {
+	ctx, cancel := monCtx()
+	defer cancel()
+	res, err := db.Col("monitors").DeleteOne(ctx, bson.M{"monitor_id": monitorID, "org_id": orgID})
+	if err != nil {
+		return err
+	}
+	// Only cascade when the org-scoped delete actually removed a monitor.
+	if res.DeletedCount == 0 {
+		return nil
+	}
+	db.Col("incidents").DeleteMany(ctx, bson.M{"monitor_id": monitorID, "org_id": orgID})
+	db.Col("monitor_rollups").DeleteMany(ctx, bson.M{"monitor_id": monitorID, "org_id": orgID})
+	return nil
+}
+
+func ListIncidents(orgID, monitorID string, limit int64) ([]models.Incident, error) {
+	ctx, cancel := monCtx()
+	defer cancel()
+	if limit <= 0 {
+		limit = 50
+	}
+	cur, err := db.Col("incidents").Find(ctx, bson.M{"monitor_id": monitorID, "org_id": orgID},
+		options.Find().SetSort(bson.M{"started_at": -1}).SetLimit(limit))
+	if err != nil {
+		return nil, err
+	}
+	var out []models.Incident
+	if err := cur.All(ctx, &out); err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
+// UptimeRollups returns hourly rollups for a monitor since the cutoff, oldest first.
+func UptimeRollups(orgID, monitorID string, since time.Time) ([]models.Rollup, error) {
+	ctx, cancel := monCtx()
+	defer cancel()
+	cur, err := db.Col("monitor_rollups").Find(ctx,
+		bson.M{"monitor_id": monitorID, "org_id": orgID, "period_start": bson.M{"$gte": since}},
+		options.Find().SetSort(bson.M{"period_start": 1}))
+	if err != nil {
+		return nil, err
+	}
+	var out []models.Rollup
+	if err := cur.All(ctx, &out); err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
+// IngestResult applies a check result to a monitor: updates state, opens/resolves
+// incidents on up<->down transitions, rolls up the hourly bucket, and fires
+// notifications on transition. Both the server scheduler and agent-reported
+// results funnel through here.
+//
+// monitorID is client-supplied on the agent path, so the caller passes the org
+// and runner it is authenticated as: orgID is the reporting agent's server org
+// and runner is its server_id. A result is only applied to a monitor owned by
+// that org and assigned to that runner. An empty orgID is rejected — it would
+// skip the ownership check entirely.
+func IngestResult(orgID, runner, monitorID string, res checker.Result) error {
+	if orgID == "" {
+		return errors.New("org id required")
+	}
+	return ingestResult(orgID, runner, monitorID, res)
+}
+
+// IngestServerScheduledResult applies a result produced by the in-process server
+// scheduler, which has no org context of its own. This is the scheduler's entry
+// point and must never be called from a request-driven path — it skips the org
+// ownership check.
+func IngestServerScheduledResult(monitorID string, res checker.Result) error {
+	return ingestResult("", models.RunnerServer, monitorID, res)
+}
+
+// ingestResult is the shared implementation. An empty orgID skips the org
+// ownership check and is only reachable via IngestServerScheduledResult.
+func ingestResult(orgID, runner, monitorID string, res checker.Result) error {
+	ctx, cancel := monCtx()
+	defer cancel()
+
+	m, err := getMonitorByID(monitorID)
+	if err != nil {
+		return err
+	}
+	// Report not-found the same way as a cross-org hit, so probing an unknown
+	// monitor_id is no quieter than probing a foreign one and stale monitors
+	// stay visible to operators.
+	if m == nil {
+		return fmt.Errorf("monitor %s not found", monitorID)
+	}
+	if orgID != "" && m.OrgID != orgID {
+		return fmt.Errorf("monitor %s belongs to another org", monitorID)
+	}
+	if m.Runner != runner {
+		return fmt.Errorf("monitor %s is not run by %s", monitorID, runner)
+	}
+
+	now := time.Now()
+	prev := m.State.Status
+	retries := m.Retries
+	if retries < 1 {
+		retries = 1
+	}
+
+	newStatus := prev
+	fails := m.State.Fails
+	if res.Up {
+		fails = 0
+		newStatus = models.StatusUp
+	} else {
+		fails++
+		if fails >= retries {
+			newStatus = models.StatusDown
+		} else if prev == "" || prev == models.StatusPending {
+			newStatus = models.StatusPending
+		}
+	}
+
+	state := bson.M{
+		"state.status":        newStatus,
+		"state.last_check_at": now,
+		"state.latency_ms":    res.LatencyMs,
+		"state.message":       res.Message,
+		"state.fails":         fails,
+	}
+	if res.CertExpiry != nil {
+		state["state.cert_expiry_at"] = *res.CertExpiry
+	}
+	if _, err := db.Col("monitors").UpdateOne(ctx, bson.M{"monitor_id": monitorID}, bson.M{"$set": state}); err != nil {
+		return err
+	}
+
+	// Hourly rollup.
+	bucket := now.Truncate(time.Hour)
+	up := 0
+	if res.Up {
+		up = 1
+	}
+	// org_id via $setOnInsert rather than the filter: a legacy bucket written
+	// before rollups were tenanted must keep accumulating, not fork in two.
+	db.Col("monitor_rollups").UpdateOne(ctx,
+		bson.M{"monitor_id": monitorID, "period_start": bucket},
+		bson.M{
+			"$inc":         bson.M{"checks": 1, "up_count": up, "sum_latency": int64(res.LatencyMs)},
+			"$setOnInsert": bson.M{"org_id": m.OrgID},
+		},
+		options.UpdateOne().SetUpsert(true))
+
+	// Transition handling.
+	if newStatus != prev {
+		switch newStatus {
+		case models.StatusDown:
+			inc := models.Incident{
+				OrgID:      m.OrgID,
+				IncidentID: uuid.NewString(),
+				MonitorID:  monitorID,
+				StartedAt:  now,
+				Cause:      res.Message,
+			}
+			db.Col("incidents").InsertOne(ctx, inc)
+			notifyTransition(m, newStatus, res.Message)
+		case models.StatusUp:
+			if prev == models.StatusDown {
+				db.Col("incidents").UpdateOne(ctx,
+					bson.M{"monitor_id": monitorID, "org_id": m.OrgID, "resolved_at": nil},
+					bson.M{"$set": bson.M{"resolved_at": now}})
+				notifyTransition(m, newStatus, res.Message)
+			}
+		}
+	}
+	return nil
+}
+
+// notifyTransition dispatches notifications on an up<->down transition to each
+// enabled channel bound to the monitor. Deliveries run in the background;
+// failures are logged, not fatal.
+func notifyTransition(m *models.Monitor, newStatus, message string) {
+	if len(m.ChannelIDs) == 0 {
+		return
+	}
+	channels, err := GetChannels(m.OrgID, m.ChannelIDs)
+	if err != nil {
+		log.Printf("notify: load channels for %s: %v", m.MonitorID, err)
+		return
+	}
+	ev := notify.Event{
+		MonitorName: m.Name,
+		Type:        m.Type,
+		OldStatus:   m.State.Status,
+		NewStatus:   newStatus,
+		Message:     message,
+		Time:        time.Now(),
+	}
+	for _, ch := range channels {
+		if !ch.Enabled {
+			continue
+		}
+		go func(c models.NotificationChannel) {
+			if err := notify.Dispatch(c, ev); err != nil {
+				log.Printf("notify: dispatch to %s (%s): %v", c.Name, c.Type, err)
+			}
+		}(ch)
+	}
+	_ = UpdateMonitor(m.OrgID, m.MonitorID, bson.M{"state.last_notified_at": time.Now()})
+}

server/internal/services/org_oidc.go

@@ -0,0 +1,52 @@
+package services
+
+import (
+	"context"
+	"time"
+
+	"github.com/mrhid6/vantage/server/internal/db"
+	"github.com/mrhid6/vantage/server/internal/models"
+	"go.mongodb.org/mongo-driver/v2/bson"
+	"go.mongodb.org/mongo-driver/v2/mongo/options"
+)
+
+func GetOrgOIDC(orgID string) (*models.OrgOIDC, error) {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	var o models.OrgOIDC
+	err := db.Col("org_oidc").FindOne(ctx, bson.M{"org_id": orgID}).Decode(&o)
+	if err != nil {
+		return nil, err
+	}
+	return &o, nil
+}
+
+func GetOrgOIDCSecret(orgID string) (string, error) {
+	o, err := GetOrgOIDC(orgID)
+	if err != nil {
+		return "", err
+	}
+	return decryptString(o.ClientSecretEnc)
+}
+
+// SaveOrgOIDC upserts the org's provider config. An empty clientSecret keeps the
+// stored secret (so the UI need not resend it).
+func SaveOrgOIDC(orgID, issuer, clientID, clientSecret string, enabled bool) error {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	set := bson.M{
+		"org_id": orgID, "issuer": issuer, "client_id": clientID,
+		"enabled": enabled, "updated_at": time.Now(),
+	}
+	if clientSecret != "" {
+		enc, err := encryptString(clientSecret)
+		if err != nil {
+			return err
+		}
+		set["client_secret_enc"] = enc
+	}
+	_, err := db.Col("org_oidc").UpdateOne(ctx,
+		bson.M{"org_id": orgID}, bson.M{"$set": set},
+		options.UpdateOne().SetUpsert(true))
+	return err
+}

server/internal/services/orgs.go

@@ -0,0 +1,168 @@
+package services
+
+import (
+	"context"
+	"fmt"
+	"log"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/mrhid6/vantage/server/internal/db"
+	"github.com/mrhid6/vantage/server/internal/models"
+	"go.mongodb.org/mongo-driver/v2/bson"
+	"go.mongodb.org/mongo-driver/v2/mongo"
+)
+
+var reservedSlugs = map[string]bool{
+	"www": true, "api": true, "app": true, "admin": true, "auth": true,
+	"install": true, "static": true, "_next": true, "default": true,
+}
+
+func GetOrg(orgID string) (*models.Org, error) {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	var o models.Org
+	err := db.Col("orgs").FindOne(ctx, bson.M{"org_id": orgID}).Decode(&o)
+	if err != nil {
+		return nil, err
+	}
+	return &o, nil
+}
+
+func GetOrgBySlug(slug string) (*models.Org, error) {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	var o models.Org
+	err := db.Col("orgs").FindOne(ctx, bson.M{"slug": slug}).Decode(&o)
+	if err != nil {
+		return nil, err
+	}
+	return &o, nil
+}
+
+// ListOrgIDs returns the org_id of every organization. Used by startup tasks
+// (e.g. seeding default workflow steps) that must run once per org.
+func ListOrgIDs() ([]string, error) {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	cursor, err := db.Col("orgs").Find(ctx, bson.M{})
+	if err != nil {
+		return nil, err
+	}
+	defer cursor.Close(ctx)
+	var orgs []models.Org
+	if err := cursor.All(ctx, &orgs); err != nil {
+		return nil, err
+	}
+	ids := make([]string, 0, len(orgs))
+	for _, o := range orgs {
+		ids = append(ids, o.OrgID)
+	}
+	return ids, nil
+}
+
+// CountOrgs returns the number of organizations on the instance. Used by
+// first-run bootstrap to tell "empty instance" from "upgraded single-tenant
+// instance whose data already sits under a migration-created org".
+func CountOrgs() (int64, error) {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	return db.Col("orgs").CountDocuments(ctx, bson.M{})
+}
+
+// FirstOrg returns the sole/earliest org. Callers must have established that
+// exactly one exists before treating it as authoritative.
+func FirstOrg() (*models.Org, error) {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	var o models.Org
+	if err := db.Col("orgs").FindOne(ctx, bson.M{}).Decode(&o); err != nil {
+		return nil, err
+	}
+	return &o, nil
+}
+
+// AdoptOrg renames an existing org to name, re-slugging it when the new slug is
+// clean to take. It exists for the upgrade path: migration 0001 stamps every
+// legacy document with the "default" org's ID, so bootstrap must claim that org
+// rather than mint a second one — otherwise the operator signs in to an empty
+// instance while all their servers and keys stay behind under "default".
+//
+// The slug is only changed when the derived one is usable and free; anything
+// else keeps the current slug, including the reserved "default", which stays
+// valid because it is pre-existing rather than newly chosen.
+func AdoptOrg(orgID, name string) (*models.Org, error) {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+
+	set := bson.M{"name": name}
+
+	slug := Slugify(name)
+	if len(slug) > 40 {
+		slug = slug[:40]
+	}
+	if len(slug) >= 3 && !reservedSlugs[slug] {
+		n, err := db.Col("orgs").CountDocuments(ctx, bson.M{"slug": slug, "org_id": bson.M{"$ne": orgID}})
+		if err != nil {
+			return nil, err
+		}
+		if n == 0 {
+			set["slug"] = slug
+		}
+	}
+
+	if _, err := db.Col("orgs").UpdateOne(ctx, bson.M{"org_id": orgID}, bson.M{"$set": set}); err != nil {
+		if mongo.IsDuplicateKeyError(err) {
+			return nil, fmt.Errorf("organization slug already taken")
+		}
+		return nil, err
+	}
+	return GetOrg(orgID)
+}
+
+func CreateOrg(name string) (*models.Org, error) {
+	base := Slugify(name)
+	if len(base) < 3 {
+		return nil, fmt.Errorf("organization name too short (slug must be >= 3 chars)")
+	}
+	if len(base) > 40 {
+		base = base[:40]
+	}
+	if reservedSlugs[base] {
+		return nil, fmt.Errorf("organization name is reserved")
+	}
+
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+
+	// Resolve slug collision by suffixing -2, -3, ...
+	slug := base
+	for i := 2; ; i++ {
+		n, err := db.Col("orgs").CountDocuments(ctx, bson.M{"slug": slug})
+		if err != nil {
+			return nil, err
+		}
+		if n == 0 {
+			break
+		}
+		slug = fmt.Sprintf("%s-%d", base, i)
+	}
+
+	o := &models.Org{OrgID: uuid.NewString(), Name: name, Slug: slug, CreatedAt: time.Now()}
+	if _, err := db.Col("orgs").InsertOne(ctx, o); err != nil {
+		if mongo.IsDuplicateKeyError(err) {
+			return nil, fmt.Errorf("organization slug already taken")
+		}
+		return nil, err
+	}
+
+	// Boot-time seeding only covers orgs that already existed, so an org created
+	// at runtime would have an empty step library until the next restart. Not
+	// fatal: the org is usable without it and seeding is retried on boot.
+	if created, updated, err := SeedDefaultSteps(o.OrgID); err != nil {
+		log.Printf("warning: failed to seed default steps for new org %s: %v", o.OrgID, err)
+	} else {
+		log.Printf("default steps seeded for new org %s: %d created, %d updated", o.OrgID, created, updated)
+	}
+	return o, nil
+}

server/internal/services/secrets.go

@@ -2,6 +2,7 @@ package services
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"sort"
 	"time"
@@ -13,25 +14,47 @@ import (
 	"go.mongodb.org/mongo-driver/v2/mongo/options"
 )
 
-// EnsureSecretIndexes creates the unique compound index on (group, key).
+// EnsureSecretIndexes creates the unique compound index on (org_id, group, key).
+// The pre-multi-tenant index was on (group, key) alone, which made a second org
+// collide on the same group/key — drop it if a live DB still carries it.
 func EnsureSecretIndexes() error {
 	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
 	defer cancel()
 
+	if err := db.Col("secrets").Indexes().DropOne(ctx, "group_1_key_1"); err != nil && !isIndexNotFound(err) {
+		return err
+	}
+
 	_, err := db.Col("secrets").Indexes().CreateOne(ctx, mongo.IndexModel{
-		Keys:    bson.D{{Key: "group", Value: 1}, {Key: "key", Value: 1}},
+		Keys:    bson.D{{Key: "org_id", Value: 1}, {Key: "group", Value: 1}, {Key: "key", Value: 1}},
 		Options: options.Index().SetUnique(true),
 	})
 	return err
 }
 
+// isIndexNotFound reports whether err is Mongo's IndexNotFound (27), returned
+// when dropping an index that was never created, or NamespaceNotFound (26),
+// returned when the collection itself does not exist yet. Both mean "there is
+// no legacy index to drop" — on a fresh install nothing has written to these
+// collections, so the drop must be tolerated or the index creation that follows
+// it never runs and a brand-new deployment crash-loops at startup.
+func isIndexNotFound(err error) bool {
+	var ce mongo.CommandError
+	if errors.As(err, &ce) {
+		return ce.Code == 27 || ce.Name == "IndexNotFound" ||
+			ce.Code == 26 || ce.Name == "NamespaceNotFound"
+	}
+	return false
+}
+
 // ListSecretGroups returns a summary of every group with its key count and
 // most recent update time.
-func ListSecretGroups() ([]models.GroupSummary, error) {
+func ListSecretGroups(orgID string) ([]models.GroupSummary, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
 	pipeline := mongo.Pipeline{
+		{{Key: "$match", Value: bson.D{{Key: "org_id", Value: orgID}}}},
 		{{Key: "$group", Value: bson.D{
 			{Key: "_id", Value: "$group"},
 			{Key: "key_count", Value: bson.D{{Key: "$sum", Value: 1}}},
@@ -68,11 +91,11 @@ func ListSecretGroups() ([]models.GroupSummary, error) {
 
 // GetSecretGroup returns the keys within a group, sorted by key name, without
 // decrypted values.
-func GetSecretGroup(group string) ([]models.Secret, error) {
+func GetSecretGroup(orgID, group string) ([]models.Secret, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
-	cursor, err := db.Col("secrets").Find(ctx, bson.M{"group": group},
+	cursor, err := db.Col("secrets").Find(ctx, bson.M{"org_id": orgID, "group": group},
 		options.Find().SetSort(bson.D{{Key: "key", Value: 1}}))
 	if err != nil {
 		return nil, err
@@ -87,9 +110,10 @@ func GetSecretGroup(group string) ([]models.Secret, error) {
 }
 
 // GetSecretGroupDecrypted returns a flat map of key → plaintext value for a
-// group. Used by the ESO read endpoint.
-func GetSecretGroupDecrypted(group string) (map[string]string, error) {
-	docs, err := GetSecretGroup(group)
+// group. Also used by the ESO read endpoint, which resolves its org from the
+// per-org bearer token rather than from a session.
+func GetSecretGroupDecrypted(orgID, group string) (map[string]string, error) {
+	docs, err := GetSecretGroup(orgID, group)
 	if err != nil {
 		return nil, err
 	}
@@ -105,12 +129,12 @@ func GetSecretGroupDecrypted(group string) (map[string]string, error) {
 }
 
 // RevealSecret returns the decrypted value of a single key.
-func RevealSecret(group, key string) (string, error) {
+func RevealSecret(orgID, group, key string) (string, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
 	var doc models.Secret
-	err := db.Col("secrets").FindOne(ctx, bson.M{"group": group, "key": key}).Decode(&doc)
+	err := db.Col("secrets").FindOne(ctx, bson.M{"org_id": orgID, "group": group, "key": key}).Decode(&doc)
 	if err == mongo.ErrNoDocuments {
 		return "", fmt.Errorf("secret not found")
 	}
@@ -121,7 +145,7 @@ func RevealSecret(group, key string) (string, error) {
 }
 
 // UpsertSecrets encrypts and writes each key/value pair into the group.
-func UpsertSecrets(group string, values map[string]string) error {
+func UpsertSecrets(orgID, group string, values map[string]string) error {
 	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
 	defer cancel()
 
@@ -131,8 +155,9 @@ func UpsertSecrets(group string, values map[string]string) error {
 			return fmt.Errorf("encrypt %s: %w", key, err)
 		}
 		_, err = db.Col("secrets").UpdateOne(ctx,
-			bson.M{"group": group, "key": key},
+			bson.M{"org_id": orgID, "group": group, "key": key},
 			bson.M{"$set": bson.M{
+				"org_id":          orgID,
 				"encrypted_value": encrypted,
 				"updated_at":      time.Now(),
 			}},
@@ -156,19 +181,19 @@ func SortedKeys(m map[string]string) []string {
 }
 
 // DeleteSecret removes a single key from a group.
-func DeleteSecret(group, key string) error {
+func DeleteSecret(orgID, group, key string) error {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
-	_, err := db.Col("secrets").DeleteOne(ctx, bson.M{"group": group, "key": key})
+	_, err := db.Col("secrets").DeleteOne(ctx, bson.M{"org_id": orgID, "group": group, "key": key})
 	return err
 }
 
 // DeleteSecretGroup removes an entire group and all its keys.
-func DeleteSecretGroup(group string) error {
+func DeleteSecretGroup(orgID, group string) error {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
-	_, err := db.Col("secrets").DeleteMany(ctx, bson.M{"group": group})
+	_, err := db.Col("secrets").DeleteMany(ctx, bson.M{"org_id": orgID, "group": group})
 	return err
 }

server/internal/services/servers.go

@@ -6,6 +6,7 @@ import (
 	"crypto/sha256"
 	"encoding/hex"
 	"fmt"
+	"log"
 	"strings"
 	"time"
 
@@ -29,13 +30,14 @@ func HashToken(token string) string {
 	return hex.EncodeToString(sum[:])
 }
 
-func CreateServer() (*models.Server, string, error) {
+func CreateServer(orgID string) (*models.Server, string, error) {
 	token, err := generateToken(32)
 	if err != nil {
 		return nil, "", err
 	}
 	expires := time.Now().Add(time.Hour)
 	s := &models.Server{
+		OrgID:         orgID,
 		ServerID:      uuid.NewString(),
 		PreRegToken:   token,
 		PreRegExpires: &expires,
@@ -52,7 +54,22 @@ func CreateServer() (*models.Server, string, error) {
 	return s, token, nil
 }
 
-func GetServer(serverID string) (*models.Server, error) {
+// GetServer looks up a server scoped to an org (handler/session use).
+func GetServer(orgID, serverID string) (*models.Server, error) {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+
+	var s models.Server
+	err := db.Col("servers").FindOne(ctx, bson.M{"server_id": serverID, "org_id": orgID}).Decode(&s)
+	if err != nil {
+		return nil, err
+	}
+	return &s, nil
+}
+
+// getServerByID looks up a server by its unique server_id with no org filter.
+// For agent/internal use only (e.g. workflow runner resolving org from a run).
+func getServerByID(serverID string) (*models.Server, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
@@ -164,6 +181,11 @@ func ValidateAgentToken(serverID, agentToken string) (*models.Server, error) {
 	if err != nil {
 		return nil, fmt.Errorf("invalid agent token")
 	}
+	// Defence in depth: every agent-path caller scopes its work by this OrgID,
+	// so a blank one would widen those queries instead of narrowing them.
+	if s.OrgID == "" {
+		return nil, fmt.Errorf("server %s has no org", serverID)
+	}
 	return &s, nil
 }
 
@@ -214,12 +236,12 @@ func UpdateServerLastSeen(serverID, agentVersion string) error {
 	return err
 }
 
-func ListServers() ([]models.Server, error) {
+func ListServers(orgID string) ([]models.Server, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
 	opts := options.Find().SetSort(bson.D{{Key: "created_at", Value: -1}})
-	cursor, err := db.Col("servers").Find(ctx, bson.M{}, opts)
+	cursor, err := db.Col("servers").Find(ctx, bson.M{"org_id": orgID}, opts)
 	if err != nil {
 		return nil, err
 	}
@@ -232,16 +254,16 @@ func ListServers() ([]models.Server, error) {
 	return servers, nil
 }
 
-func DeleteServer(serverID string) error {
+func DeleteServer(orgID, serverID string) error {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
-	_, err := db.Col("servers").DeleteOne(ctx, bson.M{"server_id": serverID})
+	_, err := db.Col("servers").DeleteOne(ctx, bson.M{"server_id": serverID, "org_id": orgID})
 	if err != nil {
 		return err
 	}
 	// Also remove assignments
-	_, err = db.Col("assignments").DeleteMany(ctx, bson.M{"server_id": serverID})
+	_, err = db.Col("assignments").DeleteMany(ctx, bson.M{"server_id": serverID, "org_id": orgID})
 	return err
 }
 
@@ -261,39 +283,73 @@ func StoreAvailableUpdates(serverID string, pkgs []models.PackageUpdate) error {
 }
 
 func MarkOfflineServers() error {
-	settings, _ := GetSettings()
-	thresholdMinutes := 5
-	if settings != nil && settings.Alerts.OfflineThresholdMinutes > 0 {
-		thresholdMinutes = settings.Alerts.OfflineThresholdMinutes
-	}
-	threshold := time.Duration(thresholdMinutes) * time.Minute
-
-	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
-	defer cancel()
-
-	cutoff := time.Now().Add(-threshold)
-
-	// Find servers about to transition to offline so we can alert on them.
-	cursor, err := db.Col("servers").Find(ctx, bson.M{
-		"status":    "active",
-		"last_seen": bson.M{"$lt": cutoff},
-	})
+	orgIDs, err := ListOrgIDs()
 	if err != nil {
 		return err
 	}
-	defer cursor.Close(ctx)
 
-	var goingOffline []models.Server
-	if err := cursor.All(ctx, &goingOffline); err != nil {
-		return err
+	// No session here, so the sweep runs per-org and each org's threshold and
+	// alert config come from that org's own settings doc. Each org gets its own
+	// deadline so a slow org can't starve the ones after it, and a failure on
+	// one org is logged rather than aborting the whole sweep.
+	for _, orgID := range orgIDs {
+		if err := markOfflineForFilter(bson.M{"org_id": orgID}, orgID); err != nil {
+			log.Printf("offline sweep failed for org %s: %v", orgID, err)
+		}
 	}
 
+	// Servers whose org_id matches no existing org (org deleted, or the doc
+	// predates the backfill) would otherwise never be swept, where the old
+	// global query caught them. Sweep them with the default threshold; there is
+	// no org settings doc to read, and no org to alert.
+	if err := markOfflineForFilter(bson.M{"org_id": bson.M{"$nin": orgIDs}}, ""); err != nil {
+		log.Printf("offline sweep failed for orphaned servers: %v", err)
+	}
+	return nil
+}
+
+// markOfflineForFilter transitions active-but-stale servers matching scope to
+// offline. orgID selects whose settings supply the threshold and alert config;
+// empty means defaults with no alerting (orphaned servers).
+func markOfflineForFilter(scope bson.M, orgID string) error {
+	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+	defer cancel()
+
+	var settings *models.Settings
+	thresholdMinutes := 5
+	if orgID != "" {
+		settings, _ = GetSettings(orgID)
+		if settings != nil && settings.Alerts.OfflineThresholdMinutes > 0 {
+			thresholdMinutes = settings.Alerts.OfflineThresholdMinutes
+		}
+	}
+	cutoff := time.Now().Add(-time.Duration(thresholdMinutes) * time.Minute)
+
+	filter := bson.M{
+		"status":    "active",
+		"last_seen": bson.M{"$lt": cutoff},
+	}
+	for k, v := range scope {
+		filter[k] = v
+	}
+
+	// Find servers about to transition to offline so we can alert on them.
+	cursor, err := db.Col("servers").Find(ctx, filter)
+	if err != nil {
+		return err
+	}
+	var goingOffline []models.Server
+	err = cursor.All(ctx, &goingOffline)
+	cursor.Close(ctx)
+	if err != nil {
+		return err
+	}
 	if len(goingOffline) == 0 {
 		return nil
 	}
 
 	for _, s := range goingOffline {
-		LogEvent("server.offline", "system", s.ServerID, "", fmt.Sprintf("%s (%s) went offline", s.Hostname, s.IPAddress))
+		LogEvent(s.OrgID, "server.offline", "system", s.ServerID, "", fmt.Sprintf("%s (%s) went offline", s.Hostname, s.IPAddress))
 		if settings != nil && settings.Alerts.Enabled && settings.Alerts.WebhookURL != "" {
 			go SendOfflineWebhook(settings.Alerts.WebhookURL, s.Hostname, s.ServerID, s.IPAddress)
 		}
@@ -302,11 +358,7 @@ func MarkOfflineServers() error {
 		}
 	}
 
-	_, err = db.Col("servers").UpdateMany(ctx,
-		bson.M{
-			"status":    "active",
-			"last_seen": bson.M{"$lt": cutoff},
-		},
+	_, err = db.Col("servers").UpdateMany(ctx, filter,
 		bson.M{"$set": bson.M{"status": "offline"}},
 	)
 	return err

server/internal/services/servers_console_test.go

@@ -1,18 +0,0 @@
-package services
-
-import "testing"
-
-func TestOSTypeFromInfo(t *testing.T) {
-	cases := map[string]string{
-		"windows amd64": "windows",
-		"linux amd64":   "linux",
-		"linux arm64":   "linux",
-		"":              "linux",
-		"darwin arm64":  "linux",
-	}
-	for in, want := range cases {
-		if got := OSTypeFromInfo(in); got != want {
-			t.Errorf("OSTypeFromInfo(%q) = %q, want %q", in, got, want)
-		}
-	}
-}

server/internal/services/settings.go

@@ -34,14 +34,47 @@ var defaultSettings = models.Settings{
 	},
 }
 
-func GetSettings() (*models.Settings, error) {
+// EnsureSettingsIndexes creates the per-org uniqueness constraints on settings.
+// Pre-multi-tenant deployments had a single global settings doc and no indexes;
+// drop any legacy index if a live DB still carries one.
+func EnsureSettingsIndexes() error {
+	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+	defer cancel()
+
+	if err := db.Col("settings").Indexes().DropOne(ctx, "secrets.read_token_hash_1"); err != nil && !isIndexNotFound(err) {
+		return err
+	}
+
+	if _, err := db.Col("settings").Indexes().CreateOne(ctx, mongo.IndexModel{
+		Keys:    bson.D{{Key: "org_id", Value: 1}},
+		Options: options.Index().SetUnique(true),
+	}); err != nil {
+		return err
+	}
+
+	// Partial so the many settings docs with no ESO token set don't collide on
+	// a missing (or empty) field. Explicitly named so it does not share Mongo's
+	// default name with the legacy index dropped above, which would make every
+	// restart drop and rebuild the enforcing index.
+	_, err := db.Col("settings").Indexes().CreateOne(ctx, mongo.IndexModel{
+		Keys: bson.D{{Key: "secrets.read_token_hash", Value: 1}},
+		Options: options.Index().SetUnique(true).SetName("settings_read_token_hash_unique").
+			SetPartialFilterExpression(bson.M{
+				"secrets.read_token_hash": bson.M{"$type": "string", "$gt": ""},
+			}),
+	})
+	return err
+}
+
+func GetSettings(orgID string) (*models.Settings, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
 	var s models.Settings
-	err := db.Col("settings").FindOne(ctx, bson.M{}).Decode(&s)
+	err := db.Col("settings").FindOne(ctx, bson.M{"org_id": orgID}).Decode(&s)
 	if err == mongo.ErrNoDocuments {
 		cp := defaultSettings
+		cp.OrgID = orgID
 		return &cp, nil
 	}
 	if err != nil {
@@ -58,7 +91,7 @@ func hashToken(token string) string {
 
 // RotateSecretsReadToken generates a new ESO read token, stores its SHA-256
 // hash, and returns the plaintext token exactly once.
-func RotateSecretsReadToken() (string, error) {
+func RotateSecretsReadToken(orgID string) (string, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
@@ -69,11 +102,14 @@ func RotateSecretsReadToken() (string, error) {
 	token := hex.EncodeToString(raw)
 
 	_, err := db.Col("settings").UpdateOne(ctx,
-		bson.M{},
-		bson.M{"$set": bson.M{
-			"secrets.read_token_hash": hashToken(token),
-			"secrets.rotated_at":      time.Now(),
-		}},
+		bson.M{"org_id": orgID},
+		bson.M{
+			"$set": bson.M{
+				"secrets.read_token_hash": hashToken(token),
+				"secrets.rotated_at":      time.Now(),
+			},
+			"$setOnInsert": bson.M{"org_id": orgID},
+		},
 		options.UpdateOne().SetUpsert(true),
 	)
 	if err != nil {
@@ -82,25 +118,33 @@ func RotateSecretsReadToken() (string, error) {
 	return token, nil
 }
 
-// VerifySecretsReadToken reports whether the supplied token matches the stored
-// hash, using a constant-time comparison.
-func VerifySecretsReadToken(token string) bool {
+// ResolveSecretsReadToken looks the presented token's hash up directly and
+// returns the owning org. This is the ESO machine-to-machine path: the org is
+// carried by the token itself, since there is no session to scope it.
+func ResolveSecretsReadToken(token string) (string, bool) {
 	if token == "" {
-		return false
+		return "", false
 	}
-	s, err := GetSettings()
-	if err != nil || s.Secrets.ReadTokenHash == "" {
-		return false
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+
+	var s models.Settings
+	err := db.Col("settings").FindOne(ctx, bson.M{"secrets.read_token_hash": hashToken(token)}).Decode(&s)
+	if err != nil || s.Secrets.ReadTokenHash == "" || s.OrgID == "" {
+		return "", false
 	}
 	expected, err := hex.DecodeString(s.Secrets.ReadTokenHash)
 	if err != nil {
-		return false
+		return "", false
 	}
 	got := sha256.Sum256([]byte(token))
-	return subtle.ConstantTimeCompare(expected, got[:]) == 1
+	if subtle.ConstantTimeCompare(expected, got[:]) != 1 {
+		return "", false
+	}
+	return s.OrgID, true
 }
 
-func SaveSettings(alerts models.AlertSettings, email models.EmailSettings, retentionDays *int) error {
+func SaveSettings(orgID string, alerts models.AlertSettings, email models.EmailSettings, retentionDays *int) error {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
@@ -116,8 +160,8 @@ func SaveSettings(alerts models.AlertSettings, email models.EmailSettings, reten
 		set["workflow_log_retention_days"] = *retentionDays
 	}
 	_, err := db.Col("settings").UpdateOne(ctx,
-		bson.M{},
-		bson.M{"$set": set},
+		bson.M{"org_id": orgID},
+		bson.M{"$set": set, "$setOnInsert": bson.M{"org_id": orgID}},
 		options.UpdateOne().SetUpsert(true),
 	)
 	return err
@@ -125,8 +169,8 @@ func SaveSettings(alerts models.AlertSettings, email models.EmailSettings, reten
 
 // GetWorkflowLogRetentionDays returns the log retention in days: 30 when unset,
 // 0 for keep-forever, or the configured value.
-func GetWorkflowLogRetentionDays() (int, error) {
-	s, err := GetSettings()
+func GetWorkflowLogRetentionDays(orgID string) (int, error) {
+	s, err := GetSettings(orgID)
 	if err != nil {
 		return 30, err
 	}

server/internal/services/stepio.go

@@ -0,0 +1,86 @@
+package services
+
+import (
+	"encoding/json"
+	"fmt"
+
+	"github.com/mrhid6/vantage/server/internal/models"
+)
+
+const StepDocKind = "vantage.step/v1"
+
+// StepDoc is the portable, id-free representation of a step.
+type StepDoc struct {
+	Kind            string              `json:"kind"`
+	Name            string              `json:"name"`
+	Description     string              `json:"description"`
+	Interpreter     string              `json:"interpreter"`
+	Script          string              `json:"script"`
+	DeclaredOutputs []string            `json:"declared_outputs"`
+	DeclaredInputs  []models.InputParam `json:"declared_inputs"`
+	SecretRefs      []string            `json:"secret_refs"`
+}
+
+// ExportStepDoc builds a portable doc from a library step (ids/source stripped).
+func ExportStepDoc(s models.WorkflowStep) StepDoc {
+	return StepDoc{
+		Kind:            StepDocKind,
+		Name:            s.Name,
+		Description:     s.Description,
+		Interpreter:     s.Interpreter,
+		Script:          s.Script,
+		DeclaredOutputs: s.DeclaredOutputs,
+		DeclaredInputs:  s.DeclaredInputs,
+		SecretRefs:      s.SecretRefs,
+	}
+}
+
+// ParseStepDoc validates a v1 doc and returns a normalized (id-free) step with
+// declared_outputs recomputed from the script.
+func ParseStepDoc(b []byte) (models.WorkflowStep, error) {
+	var d StepDoc
+	if err := json.Unmarshal(b, &d); err != nil {
+		return models.WorkflowStep{}, fmt.Errorf("invalid step JSON: %w", err)
+	}
+	if d.Kind != StepDocKind {
+		return models.WorkflowStep{}, fmt.Errorf("unsupported kind %q (want %q)", d.Kind, StepDocKind)
+	}
+	if d.Name == "" || d.Interpreter == "" {
+		return models.WorkflowStep{}, fmt.Errorf("step name and interpreter are required")
+	}
+	if d.SecretRefs == nil {
+		d.SecretRefs = []string{}
+	}
+	if d.DeclaredInputs == nil {
+		d.DeclaredInputs = []models.InputParam{}
+	}
+	return models.WorkflowStep{
+		Name:            d.Name,
+		Description:     d.Description,
+		Interpreter:     d.Interpreter,
+		Script:          d.Script,
+		DeclaredOutputs: DeriveOutputs(d.Script),
+		DeclaredInputs:  d.DeclaredInputs,
+		SecretRefs:      d.SecretRefs,
+	}, nil
+}
+
+// ImportStepToLibrary parses a doc and persists it as a new user library step.
+func ImportStepToLibrary(orgID string, b []byte) (*models.WorkflowStep, error) {
+	s, err := ParseStepDoc(b)
+	if err != nil {
+		return nil, err
+	}
+	return CreateStep(orgID, s)
+}
+
+// ExportStep loads a library step and marshals it to a portable doc.
+func ExportStep(orgID, stepID string) ([]byte, error) {
+	ctx, cancel := wfCtx()
+	defer cancel()
+	s, err := getStep(ctx, orgID, stepID)
+	if err != nil {
+		return nil, err
+	}
+	return json.MarshalIndent(ExportStepDoc(*s), "", "  ")
+}

server/internal/services/steplogs.go

@@ -2,6 +2,7 @@ package services
 
 import (
 	"bytes"
+	"log"
 	"os"
 	"path/filepath"
 	"strings"
@@ -10,6 +11,7 @@ import (
 
 	"github.com/mrhid6/vantage/server/internal/db"
 	"go.mongodb.org/mongo-driver/v2/bson"
+	"go.mongodb.org/mongo-driver/v2/mongo"
 )
 
 // WorkflowLogDir returns the base directory for workflow step logs, creating it.
@@ -164,54 +166,84 @@ func StartLogSweeper() {
 	}()
 }
 
+// sweepLogs walks the run-log dirs on disk. Log dirs are keyed by run ID, not
+// by org, and this runs with no session — so retention is resolved per run from
+// the owning org of that run's doc, with the per-org values cached for the
+// sweep. Runs whose doc is gone fall back to the default retention.
 func sweepLogs() {
-	days := retentionDays()
-	if days <= 0 {
-		return
-	}
-	cutoff := time.Now().AddDate(0, 0, -days)
 	base := WorkflowLogDir()
 	entries, err := os.ReadDir(base)
 	if err != nil {
 		return
 	}
+	cache := map[string]int{}
+	now := time.Now()
+
 	for _, e := range entries {
 		if !e.IsDir() {
 			continue
 		}
 		runID := e.Name()
 		dir := filepath.Join(base, runID)
-		if runExpired(runID, dir, cutoff) {
+
+		orgID, finishedAt, found, err := runRetentionInfo(runID)
+		if err != nil {
+			// A transient lookup failure is not evidence the run is gone —
+			// purging at the default retention here would delete logs an org
+			// had set to keep longer, or forever.
+			log.Printf("log sweep: retention lookup failed for run %s: %v", runID, err)
+			continue
+		}
+		if found && finishedAt == nil {
+			continue // still running / never finished — keep
+		}
+
+		days, ok := cache[orgID]
+		if !ok {
+			days = defaultRetentionDays
+			if orgID != "" {
+				if v, err := GetWorkflowLogRetentionDays(orgID); err == nil {
+					days = v
+				}
+			}
+			cache[orgID] = days
+		}
+		if days <= 0 {
+			continue // keep forever
+		}
+		cutoff := now.AddDate(0, 0, -days)
+
+		if found {
+			if finishedAt.Before(cutoff) {
+				_ = os.RemoveAll(dir)
+			}
+			continue
+		}
+		// run doc gone: use dir mtime
+		if fi, e := os.Stat(dir); e == nil && fi.ModTime().Before(cutoff) {
 			_ = os.RemoveAll(dir)
 		}
 	}
 }
 
-// runExpired is true when the run finished before cutoff (falling back to dir
-// mtime when the run doc is gone).
-func runExpired(runID, dir string, cutoff time.Time) bool {
+const defaultRetentionDays = 30
+
+// runRetentionInfo returns the owning org and finish time of a run, and whether
+// the run doc still exists. A non-nil error means the lookup itself failed and
+// says nothing about whether the run doc exists.
+func runRetentionInfo(runID string) (string, *time.Time, bool, error) {
 	ctx, cancel := wfCtx()
 	defer cancel()
 	var run struct {
+		OrgID      string     `bson:"org_id"`
 		FinishedAt *time.Time `bson:"finished_at"`
 	}
 	err := db.Col("workflow_runs").FindOne(ctx, bson.M{"run_id": runID}).Decode(&run)
-	if err == nil {
-		if run.FinishedAt == nil {
-			return false // still running / never finished — keep
-		}
-		return run.FinishedAt.Before(cutoff)
+	if err == mongo.ErrNoDocuments {
+		return "", nil, false, nil
 	}
-	// run doc gone: use dir mtime
-	if fi, e := os.Stat(dir); e == nil {
-		return fi.ModTime().Before(cutoff)
+	if err != nil {
+		return "", nil, false, err
 	}
-	return false
-}
-
-func retentionDays() int {
-	if v, err := GetWorkflowLogRetentionDays(); err == nil {
-		return v
-	}
-	return 30
+	return run.OrgID, run.FinishedAt, true, nil
 }

server/internal/services/stepscan.go

@@ -0,0 +1,44 @@
+package services
+
+import (
+	"regexp"
+	"strings"
+)
+
+// keyAssign matches an env-var assignment target: KEY= (captures KEY).
+var keyAssign = regexp.MustCompile(`([A-Za-z_][A-Za-z0-9_]*)=`)
+
+// DeriveOutputs scans a step script and returns the output keys it writes to
+// $WORKFLOW_ENV. Best-effort: only lines that reference WORKFLOW_ENV are
+// considered. Deduplicated, first-seen order preserved.
+func DeriveOutputs(script string) []string {
+	out := []string{}
+	seen := map[string]bool{}
+	for _, line := range strings.Split(script, "\n") {
+		if !strings.Contains(line, "WORKFLOW_ENV") {
+			continue
+		}
+		for _, m := range keyAssign.FindAllStringSubmatch(line, -1) {
+			key := m[1]
+			// Skip the sentinel itself (e.g. "WORKFLOW_ENV=..." assignments).
+			if key == "WORKFLOW_ENV" || key == "env" {
+				continue
+			}
+			if seen[key] {
+				continue
+			}
+			seen[key] = true
+			out = append(out, key)
+		}
+	}
+	return out
+}
+
+var slugStrip = regexp.MustCompile(`[^a-z0-9]+`)
+
+// Slugify converts a step name into a stable kebab-case slug.
+func Slugify(name string) string {
+	s := strings.ToLower(name)
+	s = slugStrip.ReplaceAllString(s, "-")
+	return strings.Trim(s, "-")
+}

server/internal/services/sync.go

@@ -13,7 +13,15 @@ func BuildAuthorizedKeys(serverID string) ([]string, error) {
 	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 	defer cancel()
 
+	// Agent path — no session, so the org comes from the server record itself
+	// and both follow-up queries are scoped to it.
+	srv, err := getServerByID(serverID)
+	if err != nil {
+		return nil, err
+	}
+
 	cursor, err := db.Col("assignments").Find(ctx, bson.M{
+		"org_id":     srv.OrgID,
 		"server_id":  serverID,
 		"revoked_at": nil,
 	})
@@ -30,7 +38,7 @@ func BuildAuthorizedKeys(serverID string) ([]string, error) {
 	var lines []string
 	for _, a := range assignments {
 		var key models.Key
-		err := db.Col("keys").FindOne(ctx, bson.M{"key_id": a.KeyID}).Decode(&key)
+		err := db.Col("keys").FindOne(ctx, bson.M{"key_id": a.KeyID, "org_id": srv.OrgID}).Decode(&key)
 		if err != nil {
 			continue
 		}

server/internal/services/users.go

@@ -0,0 +1,183 @@
+package services
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"strings"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/mrhid6/vantage/server/internal/db"
+	"github.com/mrhid6/vantage/server/internal/models"
+	"go.mongodb.org/mongo-driver/v2/bson"
+	"go.mongodb.org/mongo-driver/v2/mongo"
+	"golang.org/x/crypto/bcrypt"
+)
+
+// ErrLastOwner is returned when an operation would leave an org with no owner,
+// which would lock every remaining member out of org administration.
+var ErrLastOwner = errors.New("this is the organization's last owner — promote another member to owner first")
+
+// CountUsers counts users across the whole instance. It answers "is this a
+// brand new deployment", so it is deliberately unscoped; anything that asks
+// about a single tenant must use CountOrgUsers.
+func CountUsers() (int64, error) {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	return db.Col("users").CountDocuments(ctx, bson.M{})
+}
+
+func CountOrgUsers(orgID string) (int64, error) {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	return db.Col("users").CountDocuments(ctx, bson.M{"org_id": orgID})
+}
+
+// countOtherOwners counts owner-role users in the org excluding exceptUserID,
+// i.e. how many owners would remain if that user were removed or demoted.
+func countOtherOwners(orgID, exceptUserID string) (int64, error) {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	return db.Col("users").CountDocuments(ctx, bson.M{
+		"org_id":  orgID,
+		"role":    models.RoleOwner,
+		"user_id": bson.M{"$ne": exceptUserID},
+	})
+}
+
+func GetUserInOrg(orgID, userID string) (*models.User, error) {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	var u models.User
+	err := db.Col("users").FindOne(ctx, bson.M{"user_id": userID, "org_id": orgID}).Decode(&u)
+	if err != nil {
+		return nil, err
+	}
+	return &u, nil
+}
+
+func CreateUser(orgID, email, password, role, authSource string) (*models.User, error) {
+	email = strings.ToLower(strings.TrimSpace(email))
+	if email == "" {
+		return nil, fmt.Errorf("email required")
+	}
+	if !models.ValidRole(role) {
+		return nil, fmt.Errorf("invalid role %q", role)
+	}
+	u := &models.User{
+		UserID:     uuid.NewString(),
+		OrgID:      orgID,
+		Email:      email,
+		Role:       role,
+		AuthSource: authSource,
+		CreatedAt:  time.Now(),
+	}
+	if password != "" {
+		hash, err := bcrypt.GenerateFromPassword([]byte(password), 12)
+		if err != nil {
+			return nil, err
+		}
+		u.PasswordHash = string(hash)
+	}
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	if _, err := db.Col("users").InsertOne(ctx, u); err != nil {
+		if mongo.IsDuplicateKeyError(err) {
+			return nil, fmt.Errorf("email already registered")
+		}
+		return nil, err
+	}
+	return u, nil
+}
+
+func GetUserByEmail(email string) (*models.User, error) {
+	email = strings.ToLower(strings.TrimSpace(email))
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	var u models.User
+	err := db.Col("users").FindOne(ctx, bson.M{"email": email}).Decode(&u)
+	if err != nil {
+		return nil, err
+	}
+	return &u, nil
+}
+
+func VerifyPassword(u *models.User, password string) bool {
+	if u == nil || u.PasswordHash == "" {
+		return false
+	}
+	return bcrypt.CompareHashAndPassword([]byte(u.PasswordHash), []byte(password)) == nil
+}
+
+func TouchLastLogin(userID string) error {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	now := time.Now()
+	_, err := db.Col("users").UpdateOne(ctx, bson.M{"user_id": userID},
+		bson.M{"$set": bson.M{"last_login": now}})
+	return err
+}
+
+func ListUsers(orgID string) ([]models.User, error) {
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	cursor, err := db.Col("users").Find(ctx, bson.M{"org_id": orgID})
+	if err != nil {
+		return nil, err
+	}
+	defer cursor.Close(ctx)
+	var users []models.User
+	if err := cursor.All(ctx, &users); err != nil {
+		return nil, err
+	}
+	return users, nil
+}
+
+func UpdateUserRole(orgID, userID, role string) error {
+	if !models.ValidRole(role) {
+		return fmt.Errorf("invalid role %q", role)
+	}
+	target, err := GetUserInOrg(orgID, userID)
+	if err != nil {
+		return fmt.Errorf("user not found")
+	}
+	// Demoting the final owner would leave nobody able to administer the org.
+	if target.Role == models.RoleOwner && role != models.RoleOwner {
+		others, err := countOtherOwners(orgID, userID)
+		if err != nil {
+			return err
+		}
+		if others == 0 {
+			return ErrLastOwner
+		}
+	}
+
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	_, err = db.Col("users").UpdateOne(ctx,
+		bson.M{"user_id": userID, "org_id": orgID},
+		bson.M{"$set": bson.M{"role": role}})
+	return err
+}
+
+func DeleteUser(orgID, userID string) error {
+	target, err := GetUserInOrg(orgID, userID)
+	if err != nil {
+		return fmt.Errorf("user not found")
+	}
+	if target.Role == models.RoleOwner {
+		others, err := countOtherOwners(orgID, userID)
+		if err != nil {
+			return err
+		}
+		if others == 0 {
+			return ErrLastOwner
+		}
+	}
+
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+	_, err = db.Col("users").DeleteOne(ctx, bson.M{"user_id": userID, "org_id": orgID})
+	return err
+}

server/internal/services/validate.go

@@ -0,0 +1,19 @@
+package services
+
+import (
+	"fmt"
+
+	"github.com/mrhid6/vantage/server/internal/models"
+)
+
+// ValidateWorkflow checks each step ref sets exactly one of step_id / inline.
+func ValidateWorkflow(w models.Workflow) error {
+	for i, ref := range w.Steps {
+		hasLib := ref.StepID != ""
+		hasInline := ref.Inline != nil
+		if hasLib == hasInline {
+			return fmt.Errorf("step %d: exactly one of step_id or inline must be set", i)
+		}
+	}
+	return nil
+}

server/internal/services/workflow_runner.go

@@ -2,6 +2,7 @@ package services
 
 import (
 	"fmt"
+	"os"
 	"strings"
 	"time"
 
@@ -18,8 +19,8 @@ const stepDispatchGrace = 15 * time.Second
 
 // TriggerWorkflow snapshots the workflow, creates a run doc, and starts a
 // background goroutine per target server (parallel fan-out). Returns run_id.
-func TriggerWorkflow(workflowID, actor string) (string, error) {
-	wf, err := GetWorkflow(workflowID)
+func TriggerWorkflow(orgID, workflowID, actor string) (string, error) {
+	wf, err := GetWorkflow(orgID, workflowID)
 	if err != nil {
 		return "", err
 	}
@@ -29,21 +30,27 @@ func TriggerWorkflow(workflowID, actor string) (string, error) {
 	if len(wf.Steps) == 0 {
 		return "", fmt.Errorf("workflow has no steps")
 	}
+	// Re-check ownership at trigger time — targets may predate validation or a
+	// server may have been removed since the workflow was saved.
+	if err := validateTargetServers(orgID, wf.TargetServerIDs); err != nil {
+		return "", err
+	}
 
 	// Reject a concurrent run of the same workflow.
 	ctx, cancel := wfCtx()
-	running := db.Col("workflow_runs").FindOne(ctx, bson.M{"workflow_id": workflowID, "status": "running"})
+	running := db.Col("workflow_runs").FindOne(ctx, bson.M{"org_id": orgID, "workflow_id": workflowID, "status": "running"})
 	cancel()
 	if running.Err() == nil {
 		return "", fmt.Errorf("workflow already has a run in progress")
 	}
 
-	resolved, err := resolveSteps(wf)
+	resolved, err := resolveSteps(orgID, wf)
 	if err != nil {
 		return "", err
 	}
 
 	run := models.WorkflowRun{
+		OrgID:       orgID,
 		RunID:       uuid.New().String(),
 		WorkflowID:  workflowID,
 		Name:        wf.Name,
@@ -55,7 +62,7 @@ func TriggerWorkflow(workflowID, actor string) (string, error) {
 	}
 	for _, sid := range wf.TargetServerIDs {
 		hostname := sid
-		if s, e := GetServer(sid); e == nil {
+		if s, e := getServerByID(sid); e == nil {
 			hostname = s.Hostname
 		}
 		sr := models.ServerRun{ServerID: sid, Hostname: hostname, Status: "queued", RunEnv: map[string]string{}}
@@ -77,12 +84,16 @@ func TriggerWorkflow(workflowID, actor string) (string, error) {
 
 // resolveSteps freezes each workflow step ref into a ResolvedStep by loading the
 // library step and applying overrides.
-func resolveSteps(wf *models.Workflow) ([]models.ResolvedStep, error) {
+func resolveSteps(orgID string, wf *models.Workflow) ([]models.ResolvedStep, error) {
 	ctx, cancel := wfCtx()
 	defer cancel()
 	out := make([]models.ResolvedStep, 0, len(wf.Steps))
 	for _, ref := range wf.Steps {
-		lib, err := getStep(ctx, ref.StepID)
+		if ref.Inline != nil {
+			out = append(out, resolveInlineStep(ref))
+			continue
+		}
+		lib, err := getStep(ctx, orgID, ref.StepID)
 		if err != nil {
 			return nil, err
 		}
@@ -122,16 +133,45 @@ func resolveSteps(wf *models.Workflow) ([]models.ResolvedStep, error) {
 	return out, nil
 }
 
+// resolveInlineStep freezes an ad-hoc (inline) step ref into a ResolvedStep.
+func resolveInlineStep(ref models.WorkflowStepRef) models.ResolvedStep {
+	in := ref.Inline
+	inputs := map[string]string{}
+	for _, p := range in.DeclaredInputs {
+		if ref.Inputs != nil {
+			if v, ok := ref.Inputs[p.Name]; ok {
+				inputs[p.Name] = v
+				continue
+			}
+		}
+		inputs[p.Name] = p.Default
+	}
+	onFailure := ref.OnFailure
+	if onFailure == "" {
+		onFailure = "stop"
+	}
+	return models.ResolvedStep{
+		Order:       ref.Order,
+		Name:        in.Name,
+		Interpreter: in.Interpreter,
+		Script:      in.Script,
+		SecretRefs:  in.SecretRefs,
+		OnFailure:   onFailure,
+		MaxRetries:  ref.MaxRetries,
+		Inputs:      inputs,
+	}
+}
+
 // executeRun fans out one goroutine per server run and waits for all to finish.
 func executeRun(runID string) {
-	run, err := GetRun(runID)
+	run, err := getRunByID(runID)
 	if err != nil {
 		return
 	}
 	done := make(chan int, len(run.ServerRuns))
 	for i := range run.ServerRuns {
 		go func(idx int) {
-			runServer(runID, idx, run.Steps, run.ServerRuns[idx].ServerID)
+			runServer(run.OrgID, runID, idx, run.Steps, run.ServerRuns[idx].ServerID)
 			done <- idx
 		}(i)
 	}
@@ -140,7 +180,7 @@ func executeRun(runID string) {
 	}
 
 	// Aggregate status.
-	final, _ := GetRun(runID)
+	final, _ := getRunByID(runID)
 	status := "success"
 	for _, sr := range final.ServerRuns {
 		if sr.Status == "failed" {
@@ -156,7 +196,7 @@ func executeRun(runID string) {
 
 // runServer executes the resolved steps sequentially on one server, threading
 // output env forward and applying per-step failure policy.
-func runServer(runID string, srvIdx int, steps []models.ResolvedStep, serverID string) {
+func runServer(orgID, runID string, srvIdx int, steps []models.ResolvedStep, serverID string) {
 	now := time.Now()
 	setServerRun(runID, srvIdx, bson.M{"server_runs.$.status": "running", "server_runs.$.started_at": now})
 
@@ -184,13 +224,23 @@ func runServer(runID string, srvIdx int, steps []models.ResolvedStep, serverID s
 		}
 
 		// Merge secrets into command env (kept out of persisted logs).
-		secretVals := resolveSecrets(step.SecretRefs)
+		secretVals := resolveSecrets(orgID, step.SecretRefs)
 		for k, v := range secretVals {
 			allSecrets[k] = v
 		}
+		// Input values may template earlier step outputs and secrets, e.g.
+		// URL="http://example.com/$VersionNumber". Expand against runEnv (outputs
+		// threaded from prior steps) and this step's secrets before dispatch.
+		subst := map[string]string{}
+		for k, v := range runEnv {
+			subst[k] = v
+		}
+		for k, v := range secretVals {
+			subst[k] = v
+		}
 		cmdEnv := map[string]string{}
 		for k, v := range step.Inputs {
-			cmdEnv[k] = v
+			cmdEnv[k] = expandVars(v, subst)
 		}
 		for k, v := range runEnv {
 			cmdEnv[k] = v
@@ -310,7 +360,19 @@ func dispatchAndWait(serverID, commandID string, cmd *pb.RunStepCmd) *pb.StepRes
 	}
 }
 
-func resolveSecrets(refs []string) map[string]string {
+// expandVars substitutes $VAR and ${VAR} references in an input value from the
+// given lookup (prior step outputs and secrets). Unknown references expand to
+// empty, matching shell behaviour; a literal "$" is written as "$$".
+func expandVars(v string, lookup map[string]string) string {
+	return os.Expand(v, func(name string) string {
+		if name == "$" {
+			return "$"
+		}
+		return lookup[name]
+	})
+}
+
+func resolveSecrets(orgID string, refs []string) map[string]string {
 	out := map[string]string{}
 	for _, ref := range refs {
 		// ref format "group/KEY"; resolve via RevealSecret.
@@ -318,7 +380,7 @@ func resolveSecrets(refs []string) map[string]string {
 		if len(parts) != 2 {
 			continue
 		}
-		if v, err := RevealSecret(parts[0], parts[1]); err == nil {
+		if v, err := RevealSecret(orgID, parts[0], parts[1]); err == nil {
 			out[parts[1]] = v
 		}
 	}
@@ -347,7 +409,7 @@ func setServerRun(runID string, srvIdx int, set bson.M) {
 
 // serverIDAt returns the server_id at an index (positional operator needs a match).
 func serverIDAt(runID string, srvIdx int) string {
-	r, err := GetRun(runID)
+	r, err := getRunByID(runID)
 	if err != nil || srvIdx >= len(r.ServerRuns) {
 		return ""
 	}
@@ -411,7 +473,10 @@ func updateStep(runID, serverID string, order int, set bson.M) {
 
 // ---- reads ----
 
-func GetRun(runID string) (*models.WorkflowRun, error) {
+// getRunByID looks up a run by its unique run_id with no org filter. For
+// agent/internal run-execution use only (executeRun/runServer, etc.), which
+// don't have a session and instead resolve org from the run doc itself.
+func getRunByID(runID string) (*models.WorkflowRun, error) {
 	ctx, cancel := wfCtx()
 	defer cancel()
 	var r models.WorkflowRun
@@ -422,10 +487,22 @@ func GetRun(runID string) (*models.WorkflowRun, error) {
 	return &r, err
 }
 
-func ListRuns(workflowID string, limit int64) ([]models.WorkflowRun, error) {
+// GetRun looks up a run scoped to an org (handler/session use).
+func GetRun(orgID, runID string) (*models.WorkflowRun, error) {
 	ctx, cancel := wfCtx()
 	defer cancel()
-	cur, err := db.Col("workflow_runs").Find(ctx, bson.M{"workflow_id": workflowID},
+	var r models.WorkflowRun
+	err := db.Col("workflow_runs").FindOne(ctx, bson.M{"run_id": runID, "org_id": orgID}).Decode(&r)
+	if err == mongo.ErrNoDocuments {
+		return nil, fmt.Errorf("run not found")
+	}
+	return &r, err
+}
+
+func ListRuns(orgID, workflowID string, limit int64) ([]models.WorkflowRun, error) {
+	ctx, cancel := wfCtx()
+	defer cancel()
+	cur, err := db.Col("workflow_runs").Find(ctx, bson.M{"org_id": orgID, "workflow_id": workflowID},
 		options.Find().SetSort(bson.D{{Key: "started_at", Value: -1}}).SetLimit(limit))
 	if err != nil {
 		return nil, err
@@ -438,12 +515,12 @@ func ListRuns(workflowID string, limit int64) ([]models.WorkflowRun, error) {
 	return runs, nil
 }
 
-func CancelRun(runID string) error {
+func CancelRun(orgID, runID string) error {
 	now := time.Now()
 	ctx, cancel := wfCtx()
 	defer cancel()
 	_, err := db.Col("workflow_runs").UpdateOne(ctx,
-		bson.M{"run_id": runID, "status": "running"},
+		bson.M{"org_id": orgID, "run_id": runID, "status": "running"},
 		bson.M{"$set": bson.M{"status": "cancelled", "finished_at": now}})
 	return err
 }

server/internal/services/workflows.go

@@ -25,6 +25,18 @@ func EnsureWorkflowIndexes() error {
 	}); err != nil {
 		return err
 	}
+	// The pre-multi-tenant index was on slug alone, so seeding defaults for a
+	// second org collided — drop it if a live DB still carries it.
+	if err := db.Col("workflow_steps").Indexes().DropOne(ctx, "slug_1"); err != nil && !isIndexNotFound(err) {
+		return err
+	}
+	if _, err := db.Col("workflow_steps").Indexes().CreateOne(ctx, mongo.IndexModel{
+		Keys: bson.D{{Key: "org_id", Value: 1}, {Key: "slug", Value: 1}},
+		Options: options.Index().SetUnique(true).
+			SetPartialFilterExpression(bson.M{"source": "default"}),
+	}); err != nil {
+		return err
+	}
 	if _, err := db.Col("workflows").Indexes().CreateOne(ctx, mongo.IndexModel{
 		Keys: bson.D{{Key: "workflow_id", Value: 1}}, Options: options.Index().SetUnique(true),
 	}); err != nil {
@@ -38,10 +50,10 @@ func EnsureWorkflowIndexes() error {
 
 // ---- Steps ----
 
-func ListSteps() ([]models.WorkflowStep, error) {
+func ListSteps(orgID string) ([]models.WorkflowStep, error) {
 	ctx, cancel := wfCtx()
 	defer cancel()
-	cur, err := db.Col("workflow_steps").Find(ctx, bson.M{},
+	cur, err := db.Col("workflow_steps").Find(ctx, bson.M{"org_id": orgID},
 		options.Find().SetSort(bson.D{{Key: "name", Value: 1}}))
 	if err != nil {
 		return nil, err
@@ -54,14 +66,44 @@ func ListSteps() ([]models.WorkflowStep, error) {
 	return steps, nil
 }
 
-func CreateStep(s models.WorkflowStep) (*models.WorkflowStep, error) {
+// StepUsageCounts returns, per library step_id, the number of distinct
+// workflows that reference it. Inline steps have no step_id and are ignored.
+func StepUsageCounts(orgID string) (map[string]int, error) {
 	ctx, cancel := wfCtx()
 	defer cancel()
+	cur, err := db.Col("workflows").Find(ctx, bson.M{"org_id": orgID})
+	if err != nil {
+		return nil, err
+	}
+	defer cur.Close(ctx)
+	var wfs []models.Workflow
+	if err := cur.All(ctx, &wfs); err != nil {
+		return nil, err
+	}
+	counts := map[string]int{}
+	for _, w := range wfs {
+		seen := map[string]bool{}
+		for _, ref := range w.Steps {
+			if ref.StepID == "" || seen[ref.StepID] {
+				continue
+			}
+			seen[ref.StepID] = true
+			counts[ref.StepID]++
+		}
+	}
+	return counts, nil
+}
+
+func CreateStep(orgID string, s models.WorkflowStep) (*models.WorkflowStep, error) {
+	ctx, cancel := wfCtx()
+	defer cancel()
+	s.OrgID = orgID
 	s.StepID = uuid.New().String()
 	s.CreatedAt = time.Now()
 	s.UpdatedAt = s.CreatedAt
-	if s.DeclaredOutputs == nil {
-		s.DeclaredOutputs = []string{}
+	s.DeclaredOutputs = DeriveOutputs(s.Script)
+	if s.Source == "" {
+		s.Source = "user"
 	}
 	if s.SecretRefs == nil {
 		s.SecretRefs = []string{}
@@ -75,15 +117,15 @@ func CreateStep(s models.WorkflowStep) (*models.WorkflowStep, error) {
 	return &s, nil
 }
 
-func UpdateStep(stepID string, s models.WorkflowStep) error {
+func UpdateStep(orgID, stepID string, s models.WorkflowStep) error {
 	ctx, cancel := wfCtx()
 	defer cancel()
-	_, err := db.Col("workflow_steps").UpdateOne(ctx, bson.M{"step_id": stepID}, bson.M{"$set": bson.M{
+	_, err := db.Col("workflow_steps").UpdateOne(ctx, bson.M{"step_id": stepID, "org_id": orgID}, bson.M{"$set": bson.M{
 		"name":             s.Name,
 		"description":      s.Description,
 		"interpreter":      s.Interpreter,
 		"script":           s.Script,
-		"declared_outputs": s.DeclaredOutputs,
+		"declared_outputs": DeriveOutputs(s.Script),
 		"declared_inputs":  s.DeclaredInputs,
 		"secret_refs":      s.SecretRefs,
 		"updated_at":       time.Now(),
@@ -91,14 +133,14 @@ func UpdateStep(stepID string, s models.WorkflowStep) error {
 	return err
 }
 
-func DeleteStep(stepID string) error {
+func DeleteStep(orgID, stepID string) error {
 	ctx, cancel := wfCtx()
 	defer cancel()
-	if _, err := db.Col("workflow_steps").DeleteOne(ctx, bson.M{"step_id": stepID}); err != nil {
+	if _, err := db.Col("workflow_steps").DeleteOne(ctx, bson.M{"step_id": stepID, "org_id": orgID}); err != nil {
 		return err
 	}
 	// Cascade: remove this step from every workflow that references it, re-sequencing orders.
-	cur, err := db.Col("workflows").Find(ctx, bson.M{"steps.step_id": stepID})
+	cur, err := db.Col("workflows").Find(ctx, bson.M{"steps.step_id": stepID, "org_id": orgID})
 	if err != nil {
 		return err
 	}
@@ -128,9 +170,9 @@ func DeleteStep(stepID string) error {
 	return nil
 }
 
-func getStep(ctx context.Context, stepID string) (*models.WorkflowStep, error) {
+func getStep(ctx context.Context, orgID, stepID string) (*models.WorkflowStep, error) {
 	var s models.WorkflowStep
-	err := db.Col("workflow_steps").FindOne(ctx, bson.M{"step_id": stepID}).Decode(&s)
+	err := db.Col("workflow_steps").FindOne(ctx, bson.M{"step_id": stepID, "org_id": orgID}).Decode(&s)
 	if err == mongo.ErrNoDocuments {
 		return nil, fmt.Errorf("step %s not found", stepID)
 	}
@@ -139,10 +181,10 @@ func getStep(ctx context.Context, stepID string) (*models.WorkflowStep, error) {
 
 // ---- Workflows ----
 
-func ListWorkflows() ([]models.Workflow, error) {
+func ListWorkflows(orgID string) ([]models.Workflow, error) {
 	ctx, cancel := wfCtx()
 	defer cancel()
-	cur, err := db.Col("workflows").Find(ctx, bson.M{},
+	cur, err := db.Col("workflows").Find(ctx, bson.M{"org_id": orgID},
 		options.Find().SetSort(bson.D{{Key: "name", Value: 1}}))
 	if err != nil {
 		return nil, err
@@ -155,20 +197,21 @@ func ListWorkflows() ([]models.Workflow, error) {
 	return wfs, nil
 }
 
-func GetWorkflow(id string) (*models.Workflow, error) {
+func GetWorkflow(orgID, id string) (*models.Workflow, error) {
 	ctx, cancel := wfCtx()
 	defer cancel()
 	var w models.Workflow
-	err := db.Col("workflows").FindOne(ctx, bson.M{"workflow_id": id}).Decode(&w)
+	err := db.Col("workflows").FindOne(ctx, bson.M{"workflow_id": id, "org_id": orgID}).Decode(&w)
 	if err == mongo.ErrNoDocuments {
 		return nil, fmt.Errorf("workflow not found")
 	}
 	return &w, err
 }
 
-func CreateWorkflow(w models.Workflow) (*models.Workflow, error) {
+func CreateWorkflow(orgID string, w models.Workflow) (*models.Workflow, error) {
 	ctx, cancel := wfCtx()
 	defer cancel()
+	w.OrgID = orgID
 	w.WorkflowID = uuid.New().String()
 	w.CreatedAt = time.Now()
 	w.UpdatedAt = w.CreatedAt
@@ -178,16 +221,30 @@ func CreateWorkflow(w models.Workflow) (*models.Workflow, error) {
 	if w.Steps == nil {
 		w.Steps = []models.WorkflowStepRef{}
 	}
+	if err := ValidateWorkflow(w); err != nil {
+		return nil, err
+	}
+	if err := validateTargetServers(orgID, w.TargetServerIDs); err != nil {
+		return nil, err
+	}
+	normalizeInlineSteps(&w)
 	if _, err := db.Col("workflows").InsertOne(ctx, w); err != nil {
 		return nil, err
 	}
 	return &w, nil
 }
 
-func UpdateWorkflow(id string, w models.Workflow) error {
+func UpdateWorkflow(orgID, id string, w models.Workflow) error {
 	ctx, cancel := wfCtx()
 	defer cancel()
-	_, err := db.Col("workflows").UpdateOne(ctx, bson.M{"workflow_id": id}, bson.M{"$set": bson.M{
+	if err := ValidateWorkflow(w); err != nil {
+		return err
+	}
+	if err := validateTargetServers(orgID, w.TargetServerIDs); err != nil {
+		return err
+	}
+	normalizeInlineSteps(&w)
+	_, err := db.Col("workflows").UpdateOne(ctx, bson.M{"workflow_id": id, "org_id": orgID}, bson.M{"$set": bson.M{
 		"name":              w.Name,
 		"target_server_ids": w.TargetServerIDs,
 		"steps":             w.Steps,
@@ -196,9 +253,44 @@ func UpdateWorkflow(id string, w models.Workflow) error {
 	return err
 }
 
-func DeleteWorkflow(id string) error {
+// validateTargetServers rejects any target server that does not belong to the
+// org. The IDs are client-supplied and are later consumed by the runner's
+// unscoped lookups, so ownership has to be proven at the write boundary.
+func validateTargetServers(orgID string, serverIDs []string) error {
+	for _, sid := range serverIDs {
+		if _, err := GetServer(orgID, sid); err != nil {
+			return fmt.Errorf("target server %s not found", sid)
+		}
+	}
+	return nil
+}
+
+// normalizeInlineSteps derives outputs for inline steps and strips fields that
+// only belong to library steps.
+func normalizeInlineSteps(w *models.Workflow) {
+	for i := range w.Steps {
+		in := w.Steps[i].Inline
+		if in == nil {
+			continue
+		}
+		in.DeclaredOutputs = DeriveOutputs(in.Script)
+		in.StepID = ""
+		in.Slug = ""
+		in.Source = ""
+		in.CreatedAt = time.Time{}
+		in.UpdatedAt = time.Time{}
+		if in.SecretRefs == nil {
+			in.SecretRefs = []string{}
+		}
+		if in.DeclaredInputs == nil {
+			in.DeclaredInputs = []models.InputParam{}
+		}
+	}
+}
+
+func DeleteWorkflow(orgID, id string) error {
 	ctx, cancel := wfCtx()
 	defer cancel()
-	_, err := db.Col("workflows").DeleteOne(ctx, bson.M{"workflow_id": id})
+	_, err := db.Col("workflows").DeleteOne(ctx, bson.M{"workflow_id": id, "org_id": orgID})
 	return err
 }

site/.dockerignore

@@ -0,0 +1,5 @@
+node_modules
+.next
+out
+.env*
+npm-debug.log*

site/Dockerfile

@@ -0,0 +1,48 @@
+# Dependencies stage
+FROM node:26-alpine AS deps
+
+WORKDIR /app
+
+COPY package.json package-lock.json* ./
+RUN npm install
+
+# Build stage
+FROM node:26-alpine AS builder
+
+WORKDIR /app
+
+COPY --from=deps /app/node_modules ./node_modules
+COPY . .
+
+# Baked in at build time: NEXT_PUBLIC_* values are inlined into the client
+# bundle. SITE_API points at sitesvc, which serves both forms. Leave it empty
+# and contact falls back to mailto while signup reports it is unavailable.
+ARG NEXT_PUBLIC_SITE_API=""
+ARG NEXT_PUBLIC_CONTACT_EMAIL="support@hostxtra.co.uk"
+ENV NEXT_PUBLIC_SITE_API=$NEXT_PUBLIC_SITE_API
+ENV NEXT_PUBLIC_CONTACT_EMAIL=$NEXT_PUBLIC_CONTACT_EMAIL
+
+RUN npm run build
+
+# Runtime stage
+FROM node:26-alpine AS runner
+
+WORKDIR /app
+
+ENV NODE_ENV=production
+ENV NEXT_TELEMETRY_DISABLED=1
+
+RUN addgroup --system --gid 1001 nodejs && \
+    adduser --system --uid 1001 nextjs
+
+COPY --from=builder /app/public ./public
+COPY --from=builder --chown=nextjs:nodejs /app/.next/standalone ./
+COPY --from=builder --chown=nextjs:nodejs /app/.next/static ./.next/static
+
+USER nextjs
+
+EXPOSE 3000
+ENV PORT=3000
+ENV HOSTNAME="0.0.0.0"
+
+CMD ["node", "server.js"]

site/app/about/page.tsx

@@ -0,0 +1,131 @@
+import type { Metadata } from "next";
+
+export const metadata: Metadata = {
+  title: "About",
+  description:
+    "Vantage began as a weekend fix for a lost laptop and grew into a fleet control plane. How it is built, and what it deliberately does not do.",
+};
+
+export default function AboutPage() {
+  return (
+    <>
+      <section className="rail band band--open">
+        <span className="tag">About</span>
+        <h1 style={{ fontSize: "var(--s-3)", margin: "0.8rem 0 1.1rem", maxWidth: "20ch" }}>
+          Built for the fleet nobody was given a budget to manage.
+        </h1>
+
+        <div className="split" style={{ marginTop: "2.4rem" }}>
+          <div className="prose">
+            <p>
+              Vantage started as a weekend fix for a bad afternoon. A laptop was lost, and finding every server that
+              trusted its key meant SSHing into each one with a text editor open. The list lived in someone&apos;s head.
+              Two of the boxes were not on it.
+            </p>
+            <p>
+              The obvious tools were all heavier than the problem. A configuration management stack to write one file. A
+              bastion host that becomes the thing you now have to keep alive. A certificate authority with a rotation
+              story nobody wanted to own.
+            </p>
+            <p>
+              So it began with one job done properly: hold <code>authorized_keys</code> to a known state. Then the same
+              agent turned out to be the right place to run a deploy script, check whether a service was answering, and
+              open a shell when something was on fire. Each addition had to earn its place by riding the connection that
+              already existed.
+            </p>
+            <p>
+              Today it runs across homelabs, small hosting providers, and agencies who inherit client servers and need
+              to prove who can reach them.
+            </p>
+          </div>
+
+          <div>
+            <span className="tag">How it is built</span>
+            <div className="specs">
+              <div className="spec">
+                <span className="spec__k">SERVER</span>
+                <div>
+                  <h3>Go, MongoDB, Redis</h3>
+                  <p>
+                    One Go binary serving REST for the interface and gRPC for agents. MongoDB holds everything durable;
+                    Redis holds sessions and nothing else.
+                  </p>
+                </div>
+              </div>
+              <div className="spec">
+                <span className="spec__k">WEB</span>
+                <div>
+                  <h3>Next.js</h3>
+                  <p>
+                    An operations interface, not a brochure: dense tables, live log streams, and state you can read at a
+                    glance.
+                  </p>
+                </div>
+              </div>
+              <div className="spec">
+                <span className="spec__k">AGENT</span>
+                <div>
+                  <h3>Go, Linux and Windows</h3>
+                  <p>
+                    A single static binary under systemd or as a Windows service. No runtime, no dependencies, no
+                    package manager involved.
+                  </p>
+                </div>
+              </div>
+              <div className="spec">
+                <span className="spec__k">CONSOLE</span>
+                <div>
+                  <h3>Guacamole</h3>
+                  <p>
+                    Protocol handling is a solved problem. We proxy the connection and manage the credentials around it.
+                  </p>
+                </div>
+              </div>
+            </div>
+          </div>
+        </div>
+      </section>
+
+      <section className="rail band">
+        <span className="tag">Security posture</span>
+        <h2 style={{ fontSize: "var(--s-2)", marginTop: "0.7rem", maxWidth: "24ch" }}>
+          The parts worth being specific about.
+        </h2>
+        <div className="caps">
+          <article className="cap">
+            <span className="cap__k">Tokens</span>
+            <h3>Hashed, never stored plain</h3>
+            <p>
+              Agent tokens and the secrets read token are held as SHA-256 hashes. The plaintext exists on the
+              agent&apos;s own disk at 0600 and nowhere else.
+            </p>
+          </article>
+          <article className="cap">
+            <span className="cap__k">At rest</span>
+            <h3>AES-256-GCM</h3>
+            <p>
+              Private keys, passphrases, vault secrets, identity provider secrets and console credentials are encrypted
+              with a key held only by your deployment.
+            </p>
+          </article>
+          <article className="cap">
+            <span className="cap__k">One-time</span>
+            <h3>Tokens that expire and spend</h3>
+            <p>
+              Pre-registration tokens last an hour and work once. Console session tokens are consumed the moment the
+              tunnel opens.
+            </p>
+          </article>
+          <article className="cap">
+            <span className="cap__k">Recorded</span>
+            <h3>Every mutation is audited</h3>
+            <p>
+              Assignments, revocations, runs, console sessions, secret reveals and settings changes are attributed and
+              kept.
+            </p>
+          </article>
+        </div>
+      </section>
+    </>
+  );
+}

site/app/contact/page.tsx

@@ -0,0 +1,62 @@
+import type { Metadata } from "next";
+import { ContactForm } from "@/components/ContactForm";
+
+export const metadata: Metadata = {
+  title: "Contact",
+  description: "Sales questions, self-hosted licensing, security disclosures and bug reports.",
+};
+
+const CHANNELS = [
+  {
+    title: "Support",
+    body: "Everything else, including anything urgent.",
+    link: "support@hostxtra.co.uk",
+    href: "mailto:support@hostxtra.co.uk",
+  },
+  {
+    title: "Security disclosure",
+    body: "Encrypted reports, acknowledged within 72 hours.",
+    link: "support@hostxtra.co.uk",
+    href: "mailto:support@hostxtra.co.uk?subject=Security%20disclosure",
+  },
+  {
+    title: "Bugs and feature requests",
+    body: "Public tracker, read by the people who write the code.",
+    link: "git.vantage.sh/vantage",
+    href: "https://git.vantage.sh/vantage",
+  },
+  {
+    title: "Status",
+    body: "Control plane uptime and incident history.",
+    link: "status.vantage.sh",
+    href: "https://status.vantage.sh",
+  },
+];
+
+export default function ContactPage() {
+  return (
+    <section className="rail band band--open">
+      <span className="tag">Contact</span>
+      <h1 style={{ fontSize: "var(--s-3)", margin: "0.8rem 0 1rem", maxWidth: "15ch" }}>
+        Tell us what your fleet looks like.
+      </h1>
+
+      <div className="split" style={{ marginTop: "2.4rem" }}>
+        <div className="card">
+          <ContactForm />
+        </div>
+
+        <div>
+          <p className="prose">Pick the right door and you will get a faster answer.</p>
+          {CHANNELS.map((channel) => (
+            <div className="chan" key={channel.title}>
+              <h3>{channel.title}</h3>
+              <p>{channel.body}</p>
+              <a href={channel.href}>{channel.link}</a>
+            </div>
+          ))}
+        </div>
+      </div>
+    </section>
+  );
+}

site/app/layout.tsx

@@ -0,0 +1,41 @@
+import type { Metadata } from "next";
+import "./globals.css";
+import { Footer } from "@/components/Footer";
+import { Nav } from "@/components/Nav";
+import { ThemeScript } from "@/components/ThemeScript";
+
+export const metadata: Metadata = {
+  metadataBase: new URL("https://vantage.sh"),
+  title: {
+    default: "Vantage — one control plane for the whole fleet",
+    template: "%s — Vantage",
+  },
+  description:
+    "Self-hosted fleet control: SSH key assignment, workflow execution, service monitoring, a secrets vault and a browser console, across every server you manage.",
+  openGraph: {
+    type: "website",
+    siteName: "Vantage",
+    title: "Vantage — one control plane for the whole fleet",
+    description:
+      "Self-hosted fleet control: SSH keys, workflows, monitors, secrets and consoles, over one outbound agent connection.",
+  },
+  icons: { icon: "/images/vantage_logo.svg" },
+};
+
+export default function RootLayout({ children }: { children: React.ReactNode }) {
+  return (
+    <html lang="en-GB">
+      <head>
+        <ThemeScript />
+      </head>
+      <body>
+        <a className="skip" href="#main">
+          Skip to content
+        </a>
+        <Nav />
+        <main id="main">{children}</main>
+        <Footer />
+      </body>
+    </html>
+  );
+}

site/app/page.tsx

@@ -0,0 +1,210 @@
+import Link from "next/link";
+import { InstrumentPanel } from "@/components/InstrumentPanel";
+
+export default function OverviewPage() {
+  return (
+    <>
+      <div className="heroband">
+        <section className="rail hero">
+          <span className="tag">Self-hosted fleet control plane</span>
+          <h1>Your servers, under one pane of glass you actually own.</h1>
+          <p className="lede">
+            Vantage holds SSH keys, runs scripts, watches services, stores secrets and opens consoles — across every
+            machine you manage. One agent per server, outbound connections only, all state in your own database.
+          </p>
+          <div className="hero__acts">
+            <Link className="btn btn--solid" href="/start">
+              Create your organisation
+            </Link>
+            <Link className="btn btn--line" href="/platform">
+              What it does
+            </Link>
+          </div>
+          <p className="hero__foot">Free for 3 servers · Linux and Windows agents · Self-host the whole stack</p>
+        </section>
+      </div>
+
+      <InstrumentPanel />
+
+      <section className="rail band band--open">
+        <span className="tag">The problem</span>
+        <h2 style={{ fontSize: "var(--s-2)", marginTop: "0.7rem", maxWidth: "24ch" }}>
+          Six tools, six sources of truth, one afternoon lost.
+        </h2>
+        <div className="split" style={{ marginTop: "2rem" }}>
+          <p className="prose">
+            Most small fleets end up with keys in a spreadsheet, scripts in someone&apos;s home directory, uptime checks
+            in a separate service, secrets in a chat thread, and no record of who ran what. None of those systems know
+            about each other, so every question — who can reach this box, what ran on it last, is it even up — gets
+            answered by hand.
+          </p>
+          <div className="specs specs--flush">
+            <div className="spec">
+              <span className="spec__k">ONE AGENT</span>
+              <div>
+                <h3>Everything rides one connection</h3>
+                <p>
+                  Keys, steps, checks and inventory all travel over the same outbound link. Installing a second thing is
+                  not the answer.
+                </p>
+              </div>
+            </div>
+            <div className="spec">
+              <span className="spec__k">ONE RECORD</span>
+              <div>
+                <h3>Every change is an audit event</h3>
+                <p>
+                  Assignments, revocations, runs, console sessions and settings changes all land in the same log,
+                  attributed to a person.
+                </p>
+              </div>
+            </div>
+          </div>
+        </div>
+      </section>
+
+      <section className="rail band">
+        <span className="tag">Capabilities</span>
+        <h2 style={{ fontSize: "var(--s-2)", marginTop: "0.7rem", maxWidth: "22ch" }}>One control plane, six jobs.</h2>
+        <div className="caps">
+          <article className="cap">
+            <span className="cap__k">Access</span>
+            <h3>SSH keys</h3>
+            <p>
+              Assign public keys per server and revoke them softly. The agent diffs desired state against the file and
+              rewrites <code>authorized_keys</code> atomically.
+            </p>
+            <ul>
+              <li>fingerprint deduplication</li>
+              <li>agent-side keypair generation</li>
+              <li>revocation history preserved</li>
+            </ul>
+          </article>
+
+          <article className="cap">
+            <span className="cap__k">Execution</span>
+            <h3>Workflows</h3>
+            <p>
+              A library of bash and PowerShell steps with declared inputs, outputs and secret references, composed into
+              workflows that target a set of servers.
+            </p>
+            <ul>
+              <li>live streamed step logs</li>
+              <li>stop, continue or retry on failure</li>
+              <li>values passed between steps</li>
+            </ul>
+          </article>
+
+          <article className="cap">
+            <span className="cap__k">Uptime</span>
+            <h3>Monitors</h3>
+            <p>
+              HTTP, TCP, ICMP and TLS checks, run either from the control plane or from an agent inside the target
+              network.
+            </p>
+            <ul>
+              <li>incidents and uptime history</li>
+              <li>certificate expiry warnings</li>
+              <li>alerts to five channel types</li>
+            </ul>
+          </article>
+
+          <article className="cap">
+            <span className="cap__k">Secrets</span>
+            <h3>Vault</h3>
+            <p>
+              Grouped key/value secrets encrypted with AES-256-GCM, injected into workflow steps at execution and never
+              written to logs.
+            </p>
+            <ul>
+              <li>read token for External Secrets Operator</li>
+              <li>rotatable, hashed at rest</li>
+              <li>reveal is an audited action</li>
+            </ul>
+          </article>
+
+          <article className="cap">
+            <span className="cap__k">Access</span>
+            <h3>Browser console</h3>
+            <p>
+              Open an SSH, RDP or VNC session in the browser. SSH authenticates with a stored key, and every session is
+              recorded in the audit log.
+            </p>
+            <ul>
+              <li>one-time session tokens</li>
+              <li>credentials consumed on connect</li>
+              <li>no client software</li>
+            </ul>
+          </article>
+
+          <article className="cap">
+            <span className="cap__k">Health</span>
+            <h3>Inventory and updates</h3>
+            <p>
+              CPU, memory, swap, disks and kernel reported continuously, alongside pending OS package updates you can
+              apply from the interface.
+            </p>
+            <ul>
+              <li>metrics every 30 seconds</li>
+              <li>one-click package updates</li>
+              <li>agents update themselves</li>
+            </ul>
+          </article>
+        </div>
+      </section>
+
+      <section className="rail band">
+        <span className="tag">Getting started</span>
+        <h2 style={{ fontSize: "var(--s-2)", marginTop: "0.7rem", maxWidth: "20ch" }}>
+          Four steps, and the first three take a minute.
+        </h2>
+        <div className="flow">
+          <div className="flow__c">
+            <span className="flow__n">FIRST</span>
+            <h3>Create an organisation</h3>
+            <p>You become the owner. Everything inside is invisible to every other organisation.</p>
+          </div>
+          <div className="flow__c">
+            <span className="flow__n">THEN</span>
+            <h3>Add a server</h3>
+            <p>Copy the install one-liner. It expires in an hour and works exactly once.</p>
+          </div>
+          <div className="flow__c">
+            <span className="flow__n">THEN</span>
+            <h3>Assign a key</h3>
+            <p>Paste a public key and tick the servers. It lands inside 30 seconds.</p>
+          </div>
+          <div className="flow__c">
+            <span className="flow__n">AFTER</span>
+            <h3>Build from there</h3>
+            <p>Add checks, write a step, invite the team, connect your identity provider.</p>
+          </div>
+        </div>
+
+        <pre className="code" style={{ marginTop: "1.6rem" }}>
+          <i># Linux</i>
+          {"\n"}
+          <b>curl</b> -fsSL https://vantage.sh/install | bash -s -- --server-id=<b>$ID</b> --token=<b>$TOKEN</b>
+          {"\n\n"}
+          <i># Windows</i>
+          {"\n"}
+          <b>irm</b> https://vantage.sh/install.ps1 | <b>iex</b>
+        </pre>
+      </section>
+
+      <section className="rail band band--flush">
+        <div className="card card--cta">
+          <div style={{ maxWidth: "48ch" }}>
+            <h2 style={{ fontSize: "var(--s-1)" }}>Three servers, free, no card.</h2>
+            <p style={{ color: "var(--ink-2)", marginTop: "0.4rem", fontSize: "0.94rem" }}>
+              Create an organisation, install one agent, and watch a key land on a real box.
+            </p>
+          </div>
+          <Link className="btn btn--solid" href="/start">
+            Create organisation
+          </Link>
+        </div>
+      </section>
+    </>
+  );
+}

site/app/platform/page.tsx

@@ -0,0 +1,184 @@
+import type { Metadata } from "next";
+
+export const metadata: Metadata = {
+  title: "Platform",
+  description:
+    "How Vantage fits together: a control plane you run, one agent per server, and a single outbound connection between them.",
+};
+
+export default function PlatformPage() {
+  return (
+    <>
+      <section className="rail band band--open">
+        <span className="tag">Platform</span>
+        <h1 style={{ fontSize: "var(--s-3)", margin: "0.8rem 0 1rem", maxWidth: "19ch" }}>
+          How the pieces fit together.
+        </h1>
+        <p className="lede">
+          Three moving parts: a control plane you run, an agent on each server, and one outbound connection between
+          them.
+        </p>
+
+        <div className="split" style={{ marginTop: "3rem" }}>
+          <div>
+            <h2 style={{ fontSize: "var(--s-2)", maxWidth: "18ch" }}>The agent never listens.</h2>
+            <div className="prose" style={{ marginTop: "1rem" }}>
+              <p>
+                Every agent dials out to the control plane over gRPC with TLS. Nothing needs an inbound port, nothing
+                needs a static address, and a machine behind NAT is no different from one with a public IP.
+              </p>
+              <p>
+                Key state is polled on a 30-second interval, because 30 seconds is fine for access control and polling
+                is simple to reason about. Everything that should not wait — running a step, opening a console, applying
+                updates — is pushed down a bidirectional command stream the agent holds open.
+              </p>
+            </div>
+          </div>
+
+          <div className="specs specs--flush">
+            <div className="spec">
+              <span className="spec__k">POLL</span>
+              <div>
+                <h3>SyncKeys, every 30s</h3>
+                <p>The desired key set for this server. Unchanged state means no disk write at all.</p>
+              </div>
+            </div>
+            <div className="spec">
+              <span className="spec__k">PUSH</span>
+              <div>
+                <h3>Command stream</h3>
+                <p>Generate a key, run a step, apply updates, update the agent, clean up a workspace.</p>
+              </div>
+            </div>
+            <div className="spec">
+              <span className="spec__k">REPORT</span>
+              <div>
+                <h3>Inventory and checks</h3>
+                <p>
+                  Metrics every 30 seconds, a full hardware snapshot every 15 minutes, and monitor results as they
+                  complete.
+                </p>
+              </div>
+            </div>
+          </div>
+        </div>
+      </section>
+
+      <section className="rail band">
+        <span className="tag">Write path</span>
+        <h2 style={{ fontSize: "var(--s-2)", marginTop: "0.7rem", maxWidth: "24ch" }}>
+          The file is never half-written.
+        </h2>
+        <div className="split split--even" style={{ marginTop: "2rem" }}>
+          <p className="prose">
+            The agent computes the desired <code>authorized_keys</code> content, compares it to what is on disk, and
+            stops there if nothing changed. When it does need to write, it writes a temporary file in the same directory
+            and renames it over the real one. A machine that loses power mid-write keeps the file it had.
+          </p>
+          <pre className="code">
+            <i>// agent poll, simplified</i>
+            {"\n"}
+            desired := client.SyncKeys(serverID, token){"\n"}
+            current := keys.ReadAuthorizedKeys(){"\n\n"}
+            <b>if</b> !keys.StateChanged(current, desired) {"{"}
+            {"\n    "}
+            <i>// nothing to do</i>
+            {"\n    "}
+            <b>return</b> nil{"\n"}
+            {"}"}
+            {"\n\n"}
+            keys.WriteAuthorizedKeys(desired){"\n"}
+            <i>// write .tmp, os.Rename(), chmod 0600</i>
+          </pre>
+        </div>
+      </section>
+
+      <section className="rail band">
+        <span className="tag">Tenancy and identity</span>
+        <h2 style={{ fontSize: "var(--s-2)", marginTop: "0.7rem", maxWidth: "22ch" }}>
+          Organisations are the boundary.
+        </h2>
+        <div className="caps">
+          <article className="cap">
+            <span className="cap__k">Isolation</span>
+            <h3>Scoped at the query</h3>
+            <p>
+              Every server, key, workflow, monitor and secret belongs to an organisation, and every lookup is filtered
+              by it. Uniqueness constraints are enforced by the database, not by application logic.
+            </p>
+          </article>
+          <article className="cap">
+            <span className="cap__k">Roles</span>
+            <h3>Owner, admin, member</h3>
+            <p>
+              Members operate the fleet. Admins and owners manage people, identity settings and the secrets read token.
+            </p>
+          </article>
+          <article className="cap">
+            <span className="cap__k">Identity</span>
+            <h3>Local or OIDC, per organisation</h3>
+            <p>
+              Sign in with email and password, or connect your own provider. Each organisation configures its own issuer
+              and client.
+            </p>
+          </article>
+          <article className="cap">
+            <span className="cap__k">Sessions</span>
+            <h3>Server-side, 24 hours</h3>
+            <p>
+              Cookies carry an opaque identifier and nothing else. Session bodies live in Redis, so losing it signs
+              everyone out and costs no durable data.
+            </p>
+          </article>
+        </div>
+      </section>
+
+      <section className="rail band">
+        <span className="tag">What we do not build</span>
+        <h2 style={{ fontSize: "var(--s-2)", marginTop: "0.7rem", maxWidth: "22ch" }}>The scope is the feature.</h2>
+        <div className="specs" style={{ maxWidth: "70ch" }}>
+          <div className="spec">
+            <span className="spec__k">NOT A PROXY</span>
+            <div>
+              <h3>We are never in the SSH path</h3>
+              <p>
+                Vantage assigns keys; your client connects straight to the box. If our control plane is down, your SSH
+                still works.
+              </p>
+            </div>
+          </div>
+          <div className="spec">
+            <span className="spec__k">NO CUSTODY</span>
+            <div>
+              <h3>Private keys stay put by default</h3>
+              <p>
+                Keys generated on a server stay on it unless you explicitly upload the private half, and anything stored
+                is encrypted with a key only your deployment holds.
+              </p>
+            </div>
+          </div>
+          <div className="spec">
+            <span className="spec__k">NO PER-USER</span>
+            <div>
+              <h3>Root, not every account</h3>
+              <p>
+                Vantage manages one file per server. Per-user key management is a different product with a different
+                failure mode.
+              </p>
+            </div>
+          </div>
+          <div className="spec">
+            <span className="spec__k">NO PLUGINS</span>
+            <div>
+              <h3>An agent you can read in an evening</h3>
+              <p>
+                A few thousand lines of Go with no extension system. Auditability beats extensibility on a binary that
+                runs as root.
+              </p>
+            </div>
+          </div>
+        </div>
+      </section>
+    </>
+  );
+}

site/app/pricing/page.tsx

@@ -0,0 +1,154 @@
+import Link from "next/link";
+import type { Metadata } from "next";
+
+export const metadata: Metadata = {
+  title: "Pricing",
+  description:
+    "Priced per managed server. People, keys, workflows and secrets are free. Free for 3 servers, £4 per server per month, or £290 a year self-hosted.",
+};
+
+const COMPARISON: [string, string, string, string][] = [
+  ["Managed servers", "3", "Unlimited", "Unlimited"],
+  ["Members", "1", "Unlimited", "Unlimited"],
+  ["SSH key assignment", "Yes", "Yes", "Yes"],
+  ["Workflows and step library", "Yes", "Yes", "Yes"],
+  ["Monitors", "3", "Unlimited", "Unlimited"],
+  ["Secrets vault", "—", "Yes", "Yes"],
+  ["Browser console", "—", "Yes", "Yes"],
+  ["OIDC single sign-on", "—", "Yes", "Yes"],
+  ["Audit history", "30 days", "Forever", "Forever"],
+  ["Runs on your hardware", "—", "—", "Yes"],
+  ["Support", "Community", "Next business day", "Priority"],
+];
+
+export default function PricingPage() {
+  return (
+    <section className="rail band band--open">
+      <span className="tag">Pricing</span>
+      <h1 style={{ fontSize: "var(--s-3)", margin: "0.8rem 0 1rem", maxWidth: "17ch" }}>
+        Per managed server. Nothing else counts.
+      </h1>
+      <p className="lede">
+        People are free. Keys, workflows, monitors and secrets are free. You pay for servers running an agent, because
+        that is the only number that grows with you.
+      </p>
+
+      <div className="plans">
+        <div className="plan">
+          <div>
+            <div className="plan__n">Solo</div>
+            <p className="plan__d">A homelab, a couple of VPSes, and the keys on your own laptop.</p>
+          </div>
+          <div className="plan__p">
+            £0 <span>forever</span>
+          </div>
+          <ul>
+            <li>Up to 3 servers</li>
+            <li>Keys, workflows and monitors</li>
+            <li>One member, one organisation</li>
+            <li>Community support</li>
+          </ul>
+          <Link className="btn btn--line" href="/start">
+            Create organisation
+          </Link>
+        </div>
+
+        <div className="plan plan--pick">
+          <div>
+            <div className="plan__n">Fleet</div>
+            <p className="plan__d">Real infrastructure, and more than one person holding the keys.</p>
+          </div>
+          <div className="plan__p">
+            £4 <span>/ server / month</span>
+          </div>
+          <ul>
+            <li>Unlimited servers and members</li>
+            <li>Owner, admin and member roles</li>
+            <li>OIDC single sign-on</li>
+            <li>Browser console and secrets vault</li>
+            <li>Full audit history</li>
+            <li>Email support, next business day</li>
+          </ul>
+          <Link className="btn btn--solid" href="/start">
+            Start 14-day trial
+          </Link>
+        </div>
+
+        <div className="plan">
+          <div>
+            <div className="plan__n">Self-hosted</div>
+            <p className="plan__d">The whole stack on your metal, behind your own boundary.</p>
+          </div>
+          <div className="plan__p">
+            £290 <span>/ year, per install</span>
+          </div>
+          <ul>
+            <li>Everything in Fleet, no server cap</li>
+            <li>Your MongoDB, Redis and certificates</li>
+            <li>Mirror agent releases internally</li>
+            <li>Priority support and upgrade notes</li>
+          </ul>
+          <Link className="btn btn--line" href="/contact">
+            Talk to us
+          </Link>
+        </div>
+      </div>
+
+      <div className="scroll">
+        <table className="cmp">
+          <thead>
+            <tr>
+              <th>Capability</th>
+              <th>Solo</th>
+              <th>Fleet</th>
+              <th>Self-hosted</th>
+            </tr>
+          </thead>
+          <tbody>
+            {COMPARISON.map(([capability, solo, fleet, selfHosted]) => (
+              <tr key={capability}>
+                <td>{capability}</td>
+                <td>{solo}</td>
+                <td>{fleet}</td>
+                <td>{selfHosted}</td>
+              </tr>
+            ))}
+          </tbody>
+        </table>
+      </div>
+
+      <div style={{ marginTop: "3.2rem", maxWidth: "66ch" }}>
+        <span className="tag">Fine print, in plain words</span>
+        <div className="specs">
+          <div className="spec">
+            <span className="spec__k">COUNTING</span>
+            <div>
+              <h3>What counts as a server</h3>
+              <p>One running agent, one server. Remove a box and it stops billing that day.</p>
+            </div>
+          </div>
+          <div className="spec">
+            <span className="spec__k">LIMITS</span>
+            <div>
+              <h3>Going over on Solo</h3>
+              <p>
+                Nothing is deleted. A fourth agent registers and heartbeats, but stops syncing keys until you upgrade or
+                remove a server.
+              </p>
+            </div>
+          </div>
+          <div className="spec">
+            <span className="spec__k">EXIT</span>
+            <div>
+              <h3>Leaving</h3>
+              <p>
+                Export every server, key, workflow and secret group as JSON whenever you like. Agents keep their last
+                synced state on disk, so nobody is locked out mid-migration.
+              </p>
+            </div>
+          </div>
+        </div>
+      </div>
+    </section>
+  );
+}

site/app/start/page.tsx

@@ -0,0 +1,75 @@
+import type { Metadata } from "next";
+import { OrgForm } from "@/components/OrgForm";
+
+export const metadata: Metadata = {
+  title: "Create organisation",
+  description:
+    "An organisation owns its servers, keys, workflows, monitors and secrets. Free for three servers, hosted or self-hosted.",
+};
+
+export default function StartPage() {
+  return (
+    <section className="rail band band--open">
+      <div className="split">
+        <div>
+          <span className="tag">Create organisation</span>
+          <h1 style={{ fontSize: "var(--s-3)", margin: "0.8rem 0 1rem", maxWidth: "15ch" }}>
+            Set up your organisation.
+          </h1>
+          <p className="lede" style={{ fontSize: "var(--s-0)" }}>
+            An organisation owns its servers, keys, workflows, monitors and secrets. Nothing inside it is visible to any
+            other organisation. Confirm your email and it is created with you as its owner.
+          </p>
+
+          <div className="card" style={{ marginTop: "1.9rem" }}>
+            <OrgForm />
+          </div>
+        </div>
+
+        <div>
+          <span className="tag">What happens next</span>
+          <div className="specs">
+            <div className="spec">
+              <span className="spec__k">FIRST</span>
+              <div>
+                <h3>Confirm your email</h3>
+                <p>
+                  We send a link that works once. Your organisation is created when you open it, not before.
+                </p>
+              </div>
+            </div>
+            <div className="spec">
+              <span className="spec__k">THEN</span>
+              <div>
+                <h3>Add a key</h3>
+                <p>
+                  Paste the contents of <code>~/.ssh/id_ed25519.pub</code>. Vantage fingerprints it and refuses
+                  duplicates.
+                </p>
+              </div>
+            </div>
+            <div className="spec">
+              <span className="spec__k">THEN</span>
+              <div>
+                <h3>Add a server</h3>
+                <p>Run the install command as root. It expires in an hour and works once.</p>
+              </div>
+            </div>
+            <div className="spec">
+              <span className="spec__k">THEN</span>
+              <div>
+                <h3>Watch it register</h3>
+                <p>The server moves from pending to active on first sync, usually inside 30 seconds.</p>
+              </div>
+            </div>
+          </div>
+
+          <pre className="code" style={{ marginTop: "1.8rem" }}>
+            <b>curl</b> -fsSL https://vantage.sh/install | \{"\n"}
+            {"  "}bash -s -- --server-id=<b>$ID</b> --token=<b>$TOKEN</b>
+          </pre>
+        </div>
+      </div>
+    </section>
+  );
+}

site/components/ContactForm.tsx

@@ -0,0 +1,111 @@
+"use client";
+
+import { useState } from "react";
+import { Honeypot } from "@/components/Honeypot";
+import { submitContact, type SubmitResult } from "@/lib/submit";
+
+export function ContactForm() {
+  const [result, setResult] = useState<SubmitResult>({ state: "idle" });
+  const sending = result.state === "sending";
+
+  async function onSubmit(event: React.FormEvent<HTMLFormElement>) {
+    event.preventDefault();
+    const data = new FormData(event.currentTarget);
+    setResult({ state: "sending" });
+    setResult(
+      await submitContact({
+        name: String(data.get("name") ?? ""),
+        email: String(data.get("email") ?? ""),
+        servers: String(data.get("servers") ?? ""),
+        topic: String(data.get("topic") ?? ""),
+        message: String(data.get("message") ?? ""),
+        website: String(data.get("website") ?? ""),
+      })
+    );
+  }
+
+  if (result.state === "sent") {
+    return (
+      <div role="status">
+        <h2 style={{ fontSize: "var(--s-1)" }}>Message sent.</h2>
+        <p style={{ color: "var(--ink-2)", marginTop: "0.5rem" }}>
+          We reply within one business day. If it is urgent, email support@hostxtra.co.uk directly.
+        </p>
+      </div>
+    );
+  }
+
+  const fieldError = (name: string) => result.fields?.[name];
+
+  return (
+    <form className="form" onSubmit={onSubmit} noValidate>
+      <Honeypot />
+
+      <div className="field">
+        <label htmlFor="c-name">Your name</label>
+        <input id="c-name" name="name" type="text" autoComplete="name" required aria-describedby="c-name-err" />
+        {fieldError("name") && (
+          <small id="c-name-err" className="field__err">
+            {fieldError("name")}
+          </small>
+        )}
+      </div>
+
+      <div className="field">
+        <label htmlFor="c-email">Work email</label>
+        <input id="c-email" name="email" type="email" autoComplete="email" required aria-describedby="c-email-err" />
+        {fieldError("email") && (
+          <small id="c-email-err" className="field__err">
+            {fieldError("email")}
+          </small>
+        )}
+      </div>
+
+      <div className="field">
+        <label htmlFor="c-servers">Roughly how many servers?</label>
+        <select id="c-servers" name="servers" defaultValue="4–25">
+          <option>1–3</option>
+          <option>4–25</option>
+          <option>26–100</option>
+          <option>More than 100</option>
+        </select>
+      </div>
+
+      <div className="field">
+        <label htmlFor="c-topic">What is this about?</label>
+        <select id="c-topic" name="topic" defaultValue="Evaluating Vantage">
+          <option>Evaluating Vantage</option>
+          <option>Self-hosted licensing</option>
+          <option>Migrating from something else</option>
+          <option>Security disclosure</option>
+        </select>
+      </div>
+
+      <div className="field">
+        <label htmlFor="c-message">What are you trying to solve?</label>
+        <textarea
+          id="c-message"
+          name="message"
+          required
+          placeholder="We inherit client servers and can never prove who still has access…"
+          aria-describedby="c-message-err"
+        />
+        {fieldError("message") && (
+          <small id="c-message-err" className="field__err">
+            {fieldError("message")}
+          </small>
+        )}
+      </div>
+
+      <button className="btn btn--solid" type="submit" style={{ alignSelf: "flex-start" }} disabled={sending}>
+        {sending ? "Sending…" : "Send message"}
+      </button>
+
+      {result.state === "error" && result.message && (
+        <p className="field__err" role="alert">
+          {result.message}
+        </p>
+      )}
+    </form>
+  );
+}

site/components/Footer.tsx

@@ -0,0 +1,18 @@
+import Link from "next/link";
+import { NAV_LINKS } from "@/components/nav-links";
+
+export function Footer() {
+  return (
+    <footer className="foot">
+      <div className="rail foot__in">
+        <p>Vantage — self-hosted fleet control for people who own their servers.</p>
+        {NAV_LINKS.map((link) => (
+          <Link key={link.href} href={link.href}>
+            {link.label}
+          </Link>
+        ))}
+        <Link href="/start">Create organisation</Link>
+      </div>
+    </footer>
+  );
+}

site/components/Honeypot.tsx

@@ -0,0 +1,14 @@
+/*
+ * A field no person ever sees or tabs into, but an automated form-filler will
+ * happily complete. The server treats any value here as a bot. Hidden with
+ * inline styles rather than a utility class so it stays hidden even if the
+ * stylesheet fails to load.
+ */
+export function Honeypot() {
+  return (
+    <div style={{ position: "absolute", left: "-9999px", width: 1, height: 1, overflow: "hidden" }} aria-hidden="true">
+      <label htmlFor="website">Website</label>
+      <input id="website" name="website" type="text" tabIndex={-1} autoComplete="off" defaultValue="" />
+    </div>
+  );
+}

site/components/InstrumentPanel.tsx

@@ -0,0 +1,166 @@
+"use client";
+
+import { useEffect, useState } from "react";
+
+/*
+ * The hero's signature element: a fleet panel that plays one honest cycle of
+ * what the product actually does — a workflow runs three steps, a TLS monitor
+ * fails and opens an incident, a key revocation lands — then rests. It is a
+ * dramatisation, not live data, so nothing here talks to an API.
+ */
+
+type LogLine = { time: string; body: React.ReactNode };
+
+type Beat = {
+  at: number;
+  line: LogLine;
+  effect?: "incident" | "runDone" | "revoked";
+};
+
+const BEATS: Beat[] = [
+  { at: 600, line: { time: "14:22:02", body: "running · step 1/3 · pull image" } },
+  { at: 1500, line: { time: "14:22:04", body: "running · step 2/3 · migrate database" } },
+  {
+    at: 2600,
+    line: { time: "14:22:07", body: <><span className="ok">ok</span> · migrate database · exit 0</> },
+  },
+  {
+    at: 3400,
+    line: { time: "14:22:08", body: "running · step 3/3 · restart service" },
+    effect: "incident",
+  },
+  {
+    at: 4300,
+    line: { time: "14:22:10", body: <><span className="er">monitor</span> · edge-gw-02 tls · connection refused</> },
+  },
+  {
+    at: 5200,
+    line: { time: "14:22:11", body: <><span className="ok">ok</span> · restart service · exit 0</> },
+    effect: "runDone",
+  },
+  {
+    at: 6000,
+    line: { time: "14:22:12", body: <>run finished · <span className="ok">success</span> · 3 steps · 1 server</> },
+    effect: "revoked",
+  },
+];
+
+const FIRST_LINE: LogLine = { time: "14:22:01", body: "queued · deploy-app · 1 server" };
+const MAX_LINES = 7;
+
+export function InstrumentPanel() {
+  const [lines, setLines] = useState<LogLine[]>([FIRST_LINE]);
+  const [incident, setIncident] = useState(false);
+  const [runActive, setRunActive] = useState(true);
+  const [revoked, setRevoked] = useState(false);
+  const [resting, setResting] = useState(false);
+
+  useEffect(() => {
+    const reduced = window.matchMedia("(prefers-reduced-motion: reduce)").matches;
+
+    const apply = (effect: Beat["effect"]) => {
+      if (effect === "incident") setIncident(true);
+      if (effect === "runDone") setRunActive(false);
+      if (effect === "revoked") setRevoked(true);
+    };
+
+    // Reduced motion gets the finished state immediately rather than no state.
+    if (reduced) {
+      setLines([FIRST_LINE, ...BEATS.map((b) => b.line)].slice(-MAX_LINES));
+      BEATS.forEach((b) => apply(b.effect));
+      setResting(true);
+      return;
+    }
+
+    const timers = BEATS.map((beat) =>
+      window.setTimeout(() => {
+        setLines((prev) => [...prev, beat.line].slice(-MAX_LINES));
+        apply(beat.effect);
+      }, beat.at)
+    );
+    timers.push(window.setTimeout(() => setResting(true), 6600));
+
+    return () => timers.forEach(window.clearTimeout);
+  }, []);
+
+  return (
+    <div className="instrument">
+      <div className="rail">
+        <div className="instrument__bar">
+          <span>
+            <b>northgate</b> · fleet
+          </span>
+          <span>12 servers</span>
+          <span>{incident ? "10 up" : "11 up"}</span>
+          <span>{incident ? "2 down" : "1 down"}</span>
+          <span>3 monitors</span>
+          <span>{runActive ? "1 run active" : "no runs active"}</span>
+          <span className="instrument__clock">14:22:12 UTC</span>
+        </div>
+
+        <div className="panes">
+          <section className="pane" aria-label="Fleet status">
+            <h2 className="pane__h">
+              Fleet <span>{revoked ? "key revoked · 1 server updated" : "agents polling"}</span>
+            </h2>
+
+            <Row host="proxmox-node-1" sub="4 keys · 12% cpu" state="up" label="Active" />
+            <Row host="db-primary" sub="3 keys · 61% cpu" state="up" label="Active" />
+            <Row
+              host="edge-gw-02"
+              sub={incident ? "2 keys · tls refused" : "2 keys · tls 41d"}
+              state={incident ? "down" : "up"}
+              label={incident ? "Incident" : "Active"}
+            />
+            <Row host="win-build-01" sub="agent 1.4.1 · update ready" state="pend" label="Pending" />
+            <Row
+              host="app-worker-03"
+              sub={revoked ? "4 keys · 1 revoked" : "5 keys · idle"}
+              state="up"
+              label="Active"
+            />
+          </section>
+
+          <section className="pane" aria-label="Workflow run">
+            <h2 className="pane__h">
+              Run <span>run_8f31c2</span>
+            </h2>
+            <div className="stream" aria-live="polite">
+              {lines.map((line, i) => (
+                <div key={`${line.time}-${i}`}>
+                  <span className="t">{line.time}</span> {line.body}
+                </div>
+              ))}
+              {resting && (
+                <div>
+                  <span className="caret">_</span>
+                </div>
+              )}
+            </div>
+          </section>
+        </div>
+      </div>
+    </div>
+  );
+}
+
+function Row({
+  host,
+  sub,
+  state,
+  label,
+}: {
+  host: string;
+  sub: string;
+  state: "up" | "down" | "pend";
+  label: string;
+}) {
+  return (
+    <div className="frow">
+      <span className={state === "up" ? "dot" : `dot dot--${state}`} />
+      <span className="frow__host">{host}</span>
+      <span className="frow__sub">{sub}</span>
+      <span className={`chip chip--${state}`}>{label}</span>
+    </div>
+  );
+}

site/components/Logo.tsx

@@ -0,0 +1,11 @@
+// Traced from web/public/images/vantage_logo.svg. The fill is currentColor so
+// the mark follows the --logo token in both themes.
+export function Logo({ className }: { className?: string }) {
+  return (
+    <svg viewBox="246 207 533 610" className={className} aria-hidden="true" focusable="false">
+      <g transform="translate(0,1024) scale(0.1,-0.1)" fill="currentColor" stroke="none">
+        <path d="M4940 7767 c-96 -57 -528 -312 -960 -567 -678 -400 -1064 -628 -1187 -702 l-33 -20 0 -1357 0 -1357 293 -174 c160 -96 425 -252 587 -348 946 -559 1352 -799 1407 -833 34 -22 67 -39 72 -39 5 0 188 106 408 236 219 130 459 272 533 315 74 44 425 252 780 462 l645 382 0 1355 0 1355 -135 81 c-140 84 -1118 662 -1812 1070 -218 129 -402 236 -410 239 -7 3 -92 -42 -188 -98z m297 -331 c309 -182 971 -572 1208 -712 149 -88 372 -220 498 -294 l227 -135 0 -1175 0 -1175 -578 -342 c-317 -188 -694 -411 -837 -495 -143 -85 -344 -204 -447 -265 l-186 -111 -314 185 c-987 584 -1710 1013 -1725 1025 -10 8 -13 256 -13 1178 0 1099 1 1168 18 1182 9 8 150 93 312 188 162 96 417 246 565 333 805 476 1149 677 1156 677 4 0 56 -29 116 -64z M4520 6930 c-157 -93 -432 -256 -612 -362 -180 -105 -325 -195 -322 -199 2 -4 29 -21 59 -38 l55 -31 177 0 178 0 235 140 c129 77 299 177 377 222 l143 83 0 178 c0 97 -1 177 -2 177 -2 -1 -131 -77 -288 -170z M5430 6927 c0 -128 3 -177 13 -185 6 -5 77 -48 157 -95 80 -46 245 -143 366 -216 l221 -131 177 0 176 0 55 31 c30 17 57 35 60 39 3 5 -32 30 -77 56 -79 45 -673 396 -988 583 -80 47 -148 88 -152 89 -5 2 -8 -75 -8 -171z M5050 6562 c-424 -251 -560 -334 -560 -342 0 -5 33 -79 73 -165 52 -112 76 -154 87 -152 8 2 118 65 243 140 l228 135 77 -45 c251 -150 397 -233 404 -231 7 3 122 238 148 304 8 20 -15 36 -303 205 -171 101 -316 185 -321 187 -6 1 -40 -15 -76 -36z M3760 6109 c0 -6 187 -396 417 -867 229 -471 483 -994 564 -1162 l148 -305 232 0 233 0 271 560 c150 308 403 830 564 1160 160 329 291 605 291 612 0 19 -553 19 -568 1 -9 -12 -229 -480 -652 -1390 -73 -158 -136 -285 -140 -283 -3 2 -100 206 -215 452 -115 246 -291 624 -390 838 l-182 390 -287 3 c-202 2 -286 -1 -286 -9z M3270 5136 c0 -382 -3 -701 -6 -710 -4 -9 -1 -16 6 -16 9 0 136 72 278 157 l22 13 0 539 0 539 -142 82 c-79 46 -146 85 -150 87 -5 2 -8 -308 -8 -691z M6880 5779 c-47 -28 -113 -66 -147 -86 l-63 -35 0 -537 0 -538 142 -84 c78 -46 147 -85 153 -87 7 -2 10 221 10 708 0 390 -2 710 -5 710 -3 0 -43 -23 -90 -51z M3851 4935 l-1 -550 183 -107 c100 -60 268 -159 372 -222 105 -63 191 -113 193 -111 4 3 -274 573 -288 590 -6 8 -32 26 -56 40 l-44 26 0 76 0 76 -112 231 c-62 126 -143 291 -180 365 l-67 136 0 -550z M6380 5471 c0 -5 -70 -152 -156 -327 -203 -414 -194 -393 -194 -473 l0 -68 -51 -34 c-50 -34 -52 -38 -190 -323 -77 -159 -138 -290 -136 -292 2 -3 127 69 278 158 151 90 316 188 367 218 l92 54 0 548 c0 301 -2 548 -5 548 -3 0 -5 -4 -5 -9z M3722 3952 l-144 -86 49 -28 c26 -16 111 -66 188 -112 136 -80 526 -311 830 -492 83 -49 153 -91 158 -92 4 -2 6 76 5 174 l-3 178 -70 41 c-38 23 -246 146 -461 273 -216 128 -396 232 -400 231 -5 -1 -73 -40 -152 -87z M6310 4011 c-25 -16 -124 -74 -220 -131 -96 -57 -284 -168 -417 -247 l-243 -145 0 -174 c0 -96 2 -174 5 -174 2 0 37 20 77 44 40 24 195 116 343 204 149 87 369 217 490 289 121 72 241 142 268 157 26 16 47 29 47 31 0 6 -292 176 -301 174 -2 0 -24 -13 -49 -28z" />
+      </g>
+    </svg>
+  );
+}

site/components/Nav.tsx

@@ -0,0 +1,76 @@
+"use client";
+
+import Link from "next/link";
+import { usePathname } from "next/navigation";
+import { useEffect, useState } from "react";
+import { Logo } from "@/components/Logo";
+import { NAV_LINKS } from "@/components/nav-links";
+import { ThemeToggle } from "@/components/ThemeToggle";
+
+export function Nav() {
+  const pathname = usePathname();
+  const [open, setOpen] = useState(false);
+
+  // A route change should never leave the drawer hanging open behind the new page.
+  useEffect(() => {
+    setOpen(false);
+  }, [pathname]);
+
+  function current(href: string) {
+    return pathname === href || pathname === `${href}/` ? "page" : undefined;
+  }
+
+  return (
+    <>
+      <header className="nav">
+        <div className="rail nav__in">
+          <Link className="brand" href="/">
+            <Logo />
+            <b>Vantage</b>
+          </Link>
+
+          <nav className="nav__links" aria-label="Main">
+            {NAV_LINKS.map((link) => (
+              <Link key={link.href} href={link.href} aria-current={current(link.href)}>
+                {link.label}
+              </Link>
+            ))}
+          </nav>
+
+          <ThemeToggle />
+
+          <button
+            type="button"
+            className="icon-btn nav__menu"
+            aria-expanded={open}
+            aria-controls="nav-drawer"
+            onClick={() => setOpen((v) => !v)}
+          >
+            {open ? "Close" : "Menu"}
+          </button>
+
+          <Link className="btn btn--solid btn--sm nav__cta" href="/start">
+            Create organisation
+          </Link>
+        </div>
+      </header>
+
+      {open && (
+        <div className="drawer" id="nav-drawer">
+          <div className="rail">
+            <nav aria-label="Main, mobile">
+              {NAV_LINKS.map((link) => (
+                <Link key={link.href} href={link.href} aria-current={current(link.href)}>
+                  {link.label}
+                </Link>
+              ))}
+              <Link className="btn btn--solid" href="/start">
+                Create organisation
+              </Link>
+            </nav>
+          </div>
+        </div>
+      )}
+    </>
+  );
+}

site/components/OrgForm.tsx

@@ -0,0 +1,120 @@
+"use client";
+
+import { useState } from "react";
+import { Honeypot } from "@/components/Honeypot";
+import { submitSignup, type SubmitResult } from "@/lib/submit";
+
+const MIN_PASSWORD = 12;
+
+function slugify(value: string) {
+  return value
+    .toLowerCase()
+    .trim()
+    .replace(/[^a-z0-9]+/g, "-")
+    .replace(/^-|-$/g, "");
+}
+
+export function OrgForm() {
+  const [slug, setSlug] = useState("");
+  const [result, setResult] = useState<SubmitResult>({ state: "idle" });
+  const sending = result.state === "sending";
+
+  async function onSubmit(event: React.FormEvent<HTMLFormElement>) {
+    event.preventDefault();
+    const data = new FormData(event.currentTarget);
+    setResult({ state: "sending" });
+    setResult(
+      await submitSignup({
+        org_name: String(data.get("org_name") ?? ""),
+        email: String(data.get("email") ?? ""),
+        password: String(data.get("password") ?? ""),
+        website: String(data.get("website") ?? ""),
+      })
+    );
+  }
+
+  if (result.state === "sent") {
+    return (
+      <div role="status">
+        <h2 style={{ fontSize: "var(--s-1)" }}>Check your email.</h2>
+        <p style={{ color: "var(--ink-2)", marginTop: "0.5rem" }}>
+          We sent a confirmation link. Open it and <b>{slug || "your organisation"}</b> is created with you as its
+          owner. The link works once and expires in 24 hours.
+        </p>
+        <p style={{ color: "var(--ink-3)", marginTop: "0.75rem", fontSize: "0.88rem" }}>
+          Nothing exists until you confirm — if the email does not arrive, start again or contact
+          support@hostxtra.co.uk.
+        </p>
+      </div>
+    );
+  }
+
+  const fieldError = (name: string) => result.fields?.[name];
+
+  return (
+    <form className="form" onSubmit={onSubmit} noValidate>
+      <Honeypot />
+
+      <div className="field">
+        <label htmlFor="o-org">Organisation name</label>
+        <input
+          id="o-org"
+          name="org_name"
+          type="text"
+          placeholder="Northgate Systems"
+          required
+          onChange={(e) => setSlug(slugify(e.target.value))}
+          aria-describedby="o-org-err"
+        />
+        <span className="hostline">
+          <b>{slug || "your-org"}</b>.vantage.sh
+        </span>
+        {fieldError("org_name") && (
+          <small id="o-org-err" className="field__err">
+            {fieldError("org_name")}
+          </small>
+        )}
+      </div>
+
+      <div className="field">
+        <label htmlFor="o-email">Owner email</label>
+        <input id="o-email" name="email" type="email" autoComplete="email" required aria-describedby="o-email-err" />
+        <small>You become the first owner and can invite the rest of the team afterwards.</small>
+        {fieldError("email") && (
+          <small id="o-email-err" className="field__err">
+            {fieldError("email")}
+          </small>
+        )}
+      </div>
+
+      <div className="field">
+        <label htmlFor="o-pass">Password</label>
+        <input
+          id="o-pass"
+          name="password"
+          type="password"
+          autoComplete="new-password"
+          minLength={MIN_PASSWORD}
+          required
+          aria-describedby="o-pass-err"
+        />
+        <small>At least {MIN_PASSWORD} characters. Use a manager — you are about to manage SSH keys with it.</small>
+        {fieldError("password") && (
+          <small id="o-pass-err" className="field__err">
+            {fieldError("password")}
+          </small>
+        )}
+      </div>
+
+      <button className="btn btn--solid" type="submit" style={{ alignSelf: "flex-start" }} disabled={sending}>
+        {sending ? "Sending…" : "Create organisation"}
+      </button>
+
+      {result.state === "error" && result.message && (
+        <p className="field__err" role="alert">
+          {result.message}
+        </p>
+      )}
+    </form>
+  );
+}

site/components/ThemeScript.tsx

@@ -0,0 +1,16 @@
+// Applies the stored theme before first paint. Without this the page renders in
+// the OS theme for a frame and then snaps to the stored one.
+const script = `
+(function(){
+  try {
+    var t = localStorage.getItem("vantage-theme");
+    if (t === "light" || t === "dark") {
+      document.documentElement.setAttribute("data-theme", t);
+    }
+  } catch (e) {}
+})();
+`;
+
+export function ThemeScript() {
+  return <script dangerouslySetInnerHTML={{ __html: script }} />;
+}

site/components/ThemeToggle.tsx

@@ -0,0 +1,38 @@
+"use client";
+
+import { useEffect, useState } from "react";
+
+const STORAGE_KEY = "vantage-theme";
+
+type Theme = "light" | "dark";
+
+function currentTheme(): Theme {
+  const set = document.documentElement.getAttribute("data-theme");
+  if (set === "light" || set === "dark") return set;
+  return window.matchMedia("(prefers-color-scheme: dark)").matches ? "dark" : "light";
+}
+
+export function ThemeToggle() {
+  const [theme, setTheme] = useState<Theme | null>(null);
+
+  useEffect(() => {
+    setTheme(currentTheme());
+  }, []);
+
+  function toggle() {
+    const next: Theme = currentTheme() === "dark" ? "light" : "dark";
+    document.documentElement.setAttribute("data-theme", next);
+    window.localStorage.setItem(STORAGE_KEY, next);
+    setTheme(next);
+  }
+
+  // Until mounted the rendered label would disagree with the server output, so
+  // the button carries a neutral label on first paint.
+  const label = theme === null ? "Theme" : theme === "dark" ? "Light" : "Dark";
+
+  return (
+    <button type="button" className="icon-btn" onClick={toggle} aria-label={`Switch to ${label.toLowerCase()} theme`}>
+      {label}
+    </button>
+  );
+}

site/components/nav-links.ts

@@ -0,0 +1,6 @@
+export const NAV_LINKS = [
+  { href: "/platform", label: "Platform" },
+  { href: "/pricing", label: "Pricing" },
+  { href: "/about", label: "About" },
+  { href: "/contact", label: "Contact" },
+];

site/lib/submit.ts

@@ -0,0 +1,87 @@
+/*
+ * The marketing site is a static bundle. Both forms post to sitesvc, which owns
+ * the contact mailer and the signup flow; the control plane is not involved.
+ *
+ * The base URL is baked in at build time. Contact falls back to composing an
+ * email when sitesvc is not configured, so it never silently swallows what
+ * someone typed. Signup has no fallback: an account cannot be created over
+ * mailto, so the form says so rather than pretending.
+ */
+
+const SITE_API = (process.env.NEXT_PUBLIC_SITE_API ?? "").replace(/\/$/, "");
+const FALLBACK_ADDRESS = process.env.NEXT_PUBLIC_CONTACT_EMAIL ?? "support@hostxtra.co.uk";
+
+export type SubmitState = "idle" | "sending" | "sent" | "error";
+
+export type SubmitResult = {
+  state: SubmitState;
+  /** Message to show when the submission was refused. */
+  message?: string;
+  /** Per-field messages, keyed by field name. */
+  fields?: Record<string, string>;
+};
+
+type FieldProblem = { field: string; message: string };
+
+async function post(url: string, payload: unknown): Promise<SubmitResult> {
+  try {
+    const res = await fetch(url, {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify(payload),
+    });
+
+    if (res.ok) return { state: "sent" };
+
+    const data = await res.json().catch(() => null);
+    const problems: FieldProblem[] = data?.fields ?? [];
+
+    return {
+      state: "error",
+      message: data?.error ?? "That did not go through. Try again in a moment.",
+      fields: Object.fromEntries(problems.map((p) => [p.field, p.message])),
+    };
+  } catch {
+    return { state: "error", message: "We could not reach the server. Check your connection and try again." };
+  }
+}
+
+export async function submitContact(fields: {
+  name: string;
+  email: string;
+  servers: string;
+  topic: string;
+  message: string;
+  website: string;
+}): Promise<SubmitResult> {
+  if (!SITE_API) {
+    const body = [
+      `Name: ${fields.name}`,
+      `Email: ${fields.email}`,
+      `Servers: ${fields.servers}`,
+      `Topic: ${fields.topic}`,
+      "",
+      fields.message,
+    ].join("\n");
+    window.location.href = `mailto:${FALLBACK_ADDRESS}?subject=${encodeURIComponent(
+      `Vantage enquiry — ${fields.topic}`
+    )}&body=${encodeURIComponent(body)}`;
+    return { state: "sent" };
+  }
+  return post(`${SITE_API}/api/contact`, fields);
+}
+
+export async function submitSignup(fields: {
+  org_name: string;
+  email: string;
+  password: string;
+  website: string;
+}): Promise<SubmitResult> {
+  if (!SITE_API) {
+    return {
+      state: "error",
+      message: `Signup is not available from here yet. Email ${FALLBACK_ADDRESS} and we will set you up.`,
+    };
+  }
+  return post(`${SITE_API}/api/signup`, fields);
+}

site/next-env.d.ts

@@ -0,0 +1,6 @@
+/// <reference types="next" />
+/// <reference types="next/image-types/global" />
+import "./.next/types/routes.d.ts";
+
+// NOTE: This file should not be edited
+// see https://nextjs.org/docs/app/api-reference/config/typescript for more information.