Merge pull request #14 from Dvorinka/feat/personal-access-tokens

feat: personal access tokens with scoped bearer auth

.github/FUNDING.yml

@@ -0,0 +1 @@
+github: Dvorinka

README.md

@@ -56,7 +56,7 @@ dedicated console.
 - **Template catalog** — 30 official templates (NocoDB, Plausible, MinIO, n8n, …) plus user-defined templates with JSON upload and owner-scoped edit/delete
 - **Metrics** — per-service Docker stats plus host CPU/memory/disk telemetry
 - **HA & scaling** — failover policies, node agents, autoscaling rules
-- **Auth & security** — Better Auth sessions (email/password self-hosted), audit logs, vulnerability scans
+- **Auth & security** — Better Auth sessions (email/password self-hosted), personal access tokens with `read`/`write`/`admin` scopes for CLI and agents, audit logs, vulnerability scans
 - **Integrated docs** — searchable documentation rendered in-app on the landing page and `/docs`; syncs from GitHub, falls back to a bundled snapshot offline
 - **Networking** — Traefik reverse proxy, optional Cloudflare Tunnel
 

ROADMAP2.md

@@ -0,0 +1,396 @@
+# Containr — Roadmap 2: The dflow Convergence
+
+Last verified: **2026-10-08** — deep pass over the full dflow codebase
+(`~/Desktop/PROG+HTML/dflow`, community fork of dflow-sh/dflow, frozen for
+preservation) against Containr `main` (`15b5b0e`).
+
+Goal: make Containr the superior successor — absorb everything worth having
+from dflow, then go past it and past Railway. Same power for humans and
+agents: the UI is the friendly face, the CLI and MCP are the machine face,
+all three sit on the same API. **Branding stays Containr/Vertice** —
+white-labeling is an admin option, not an identity change.
+
+---
+
+## 1. What dflow actually is (verified)
+
+Next.js 15 + PayloadCMS 3 + MongoDB + Redis/BullMQ. Runtime model: every
+host is a remote server driven **over SSH** with Dokku auto-installed as the
+per-app container manager. Async work = 43 BullMQ queue processors
+(`src/queues/**`), live logs via Redis pub/sub → SSE endpoint, persisted to
+deployment records. ~30 collections, ~180 server actions, 11 dashboard
+routes per service (general, deployments, logs, environment, domains,
+volumes, proxy, scaling, backups, settings).
+
+### Full capability matrix — dflow → Containr
+
+Legend: ✅ done · 🟡 partial/different · ❌ missing · ➖ not applicable (dokku/SSH-specific)
+
+#### Deployment pipeline
+| dflow capability | Containr status | Action |
+|---|---|---|
+| Queued deployments, serialized per server, cancellable while queued | ❌ goroutine fire-and-forget; `CancelDeployment` exists in engine but is **never routed** | Add deploy queue + cancel route |
+| Deployment states `queued → building → success/failed`, live SSE logs + persisted record | 🟡 states exist (`building→running`); WS streams builds; no queued state | Extend status + stream |
+| Redeploy cache modes (`ps:rebuild` vs fresh pull); webhook pushes force `no-cache` | ❌ always rebuilds | Add `cache` flag on deploy/redeploy |
+| Cancel queued deployment from UI | ❌ (engine fn unrouted) | Route it |
+| Per-service job ordering | ❌ concurrent deploys of one service can race | Serialize per service |
+
+#### Service configuration
+| dflow capability | Containr status | Action |
+|---|---|---|
+| Multi-domain per service + default flag + cert type | ❌ single `services.domain` | `service_domains` table |
+| Cert types: letsencrypt / none / Cloudflare Origin / **CF-for-SaaS custom hostnames** | 🟡 Traefik ACME only | LE default; CF custom hostnames deferred (only needed if hosting tenant domains) |
+| DNS preflight check (`checkDNSConfigAction`) | ❌ | `GET /services/:id/domains/check` |
+| Volumes per service + dangling-volume browser/attach/delete | ❌ engine supports `VolumeMounts`; no API/UI | `services.volumes` jsonb + Volumes tab |
+| Resource **limits and reservations**, per-server defaults | 🟡 `cpu_limit`/`memory_limit` int columns (unclear units) | explicit string units + reserves + instance defaults |
+| Pre-deploy capacity check (dokku binary interceptor) | ❌ | Node capacity check pre-deploy/scale |
+| Maintenance mode toggle | ❌ | Traefik middleware label |
+| Basic-auth gate per service (`toggleHttpAuth`) | ❌ | Traefik basic-auth middleware |
+| Raw proxy config override (`setServiceNginxConfig`) | ❌ | Allowlisted `traefik_labels` jsonb |
+| Service clone (full config copy) | ❌ | `POST /services/:id/clone` |
+| Move service between projects (`SwitchServiceProjectDialog`) | ❌ | `PUT /services/:id {project_id}` |
+| Env var save → hot-apply + optional restart (`noRestart`) | ❌ rows rewritten; containers keep stale env until manual redeploy | Apply→redeploy flow |
+| Static-site builder; Heroku buildpacks | 🟡 railpack/nixpacks/dockerfile auto-detect, no user override | `services.builder` enum + static builder |
+| Docker registry accounts (private registries) | ❌ | `registry_credentials` per service/instance |
+
+#### Databases
+| dflow capability | Containr status | Action |
+|---|---|---|
+| Managed DBs (pg/mysql/mariadb/mongo/redis/clickhouse via dokku plugins) | ✅ plus dragonfly | — |
+| **External DB registration** (Neon/Atlas/RDS conn details + test) | ❌ | `provider=external` + test endpoint |
+| Internal backups + restore (same-type enforced) | ✅ | — |
+| Scheduled backups | ✅ cron field | — |
+| **External S3 backups** (auth/schedule/unschedule ops) | ❌ | `backup_targets` + shipper |
+| **Cross-server DB migration** (export → sftp hop → import) | ❌ | `POST /databases/:id/migrate` |
+| Public DB port expose/unexpose | ❌ loopback only | opt-in TCP expose |
+| DB → service env linking (`dokku link`) | ✅ Bind action injects `DATABASE_URL` | — |
+
+#### Servers / infrastructure
+| dflow capability | Containr status | Action |
+|---|---|---|
+| Attach server = paste IP + SSH key; Dokku auto-installs | 🟡 agents exist, manual install | **SSH bootstrap**: one SSH session installs agent, never needed again |
+| Tailscale/Headscale, NetBird, ZeroTier enroll + mesh connect | ❌ | Agent mesh-IP detection first; auth-key enroll optional |
+| AWS EC2 + Hetzner provisioning w/ cloud-init | ❌ | IaC (ROADMAP §2.2): Terraform Proxmox/AWS + Hetzner module |
+| Security groups (AWS SG rules, sync) | ❌ | Under IaC provider modules |
+| Server-level global/wildcard domain (`updateServerDomain`) | ❌ | `nodes.default_domain` |
+| Scheduled server cleanup (dokku cleanup → docker prune) | ❌ | Agent `prune` command + schedule |
+| Queue flush/inspect per server (`flushServerQueues`, `getServerQueues`) | ➖ no queue yet | Ops page shows jobs once queue exists |
+| Arbitrary SSH command on server (`executeCommand`) | ➖ exec is container-scoped | Node terminal via agent (PTY) |
+| Server onboarding wizard, reset, cleanup-on-delete | 🟡 token issue/revoke only | Onboarding flow + `nodes add` UX |
+| Ansible playbooks (stored, runnable per server, exec history) | ➖ | **Defer** — security surface; SSH bootstrap covers provisioning |
+| Per-node resource dashboards | 🟡 heartbeat telemetry exists | Node detail page |
+
+#### Platform / org
+| dflow capability | Containr status | Action |
+|---|---|---|
+| Multi-tenant organisations + custom RBAC (per-resource CRUD, create-limits, read-scope) | 🟡 `is_admin` + `project_members` | Defer full RBAC; design PAT scopes role-compatible |
+| Team invite links/emails | 🟡 manual user create only | `POST /invites` → link join |
+| User API keys → full REST API | ❌ **the big gap** — session/JWT only | PATs (Phase A) |
+| Outbound webhooks (collection events, HMAC, custom headers) | ❌ | `webhooks` table + delivery log |
+| Banners (global + per-tenant, dismissible) | ❌ | `banners` table + top bar |
+| White-label (logo/favicon/title/OG/theme colors) | ❌ | `app_settings.branding` — **Containr remains the default brand** |
+| Activity feed (icon/severity/category/metadata, 90-day TTL job) | 🟡 `audit_logs` coarser | Enrich + project feed UI |
+| Admin metrics dashboard (users/servers/queued/failed counts) | 🟡 `/admin` exists | Extend |
+| **Impersonate user** | ❌ | Admin support tool |
+| In-app docs (content-collections markdown) | ✅ `/docs` | — |
+| Dynamic client-side filter framework (535-line `filter.utils`) | 🟡 audit page filters only | Reuse pattern for tables |
+| Ops "bubble" (queue monitor, terminal, sync) | ❌ | Operations page |
+
+#### Agent surface
+| dflow capability | Containr status | Action |
+|---|---|---|
+| Payload REST/GraphQL auto-API + per-user API keys | ❌ | PATs + OpenAPI (better: typed, documented) |
+| CLI | ❌ | none in dflow — **Containr CLI is a genuine leap past it** |
+| MCP | ❌ | none — same leap |
+| Railway migrator script | ❌ | `containr migrate railway` |
+| GitHub App **manifest flow** (instance self-provisions the App) | 🟡 env-configured App | `POST /git/github-app/manifest` flow |
+| fork-sync webhook deploy trigger | ❌ | ~20 lines on receiver |
+| Azure DevOps provider | ❌ | Optional (dflow does token-clone only) |
+
+---
+
+## 2. The spine: agent-first platform
+
+*Agent and user on the same level.* Concretely:
+
+- Every operation reachable by a **personal access token**, not just
+  sessions.
+- Every operation in the **CLI** with `--json`, stable exit codes, `--yes`.
+- Every operation in an **MCP server** (stdio first) so Devin/Claude Code/
+  Cursor/Codex drive Containr natively.
+- `openapi.yaml` stays the single source of truth; UI, CLI, MCP are thin
+  clients.
+
+### Phase A — Agent surface (do this FIRST; everything after inherits it)
+
+- [ ] **Personal access tokens.** `user_tokens` table — `id`, `user_id`,
+  `name`, `key_prefix` (`cnp_…`), `key_hash` (sha256), `scopes`
+  (`read`/`write`/`admin`, later resource-scoped), `expires_at`,
+  `last_used_at`. Model on `agent_auth_tokens` (raw shown once, hash stored).
+  `POST/GET/DELETE /api/v1/user/tokens`; `middleware.Auth` accepts
+  `Bearer cnp_…`; Settings → Access tokens UI with one-time display +
+  copyable CLI/MCP config snippets.
+- [ ] **CLI v2.** Current `cmd/cli` is auth + projects only. Full client:
+  - `auth login [--token] [--url] [--profile]` (multi-instance profiles).
+  - Resources: `projects services deploy logs exec variables databases
+    cron templates nodes scaling ha security gateway notifications
+    webhooks domains volumes backups admin`.
+  - `containr up` / `deploy` — cwd→service create-or-update→watch
+    (the `railway up` equivalent; single most important command).
+  - `containr link` (bind cwd→service), `containr open`, `containr run`,
+    `containr shell` (PTY once exec upgrade lands), `completion`.
+  - Universal `--json`, `-o yaml`, `--watch`, `--yes`; exit codes
+    0/1/2/3/4 (ok/err/not-found/unauthorized/conflict).
+  - GoReleaser → GH releases + `install.sh`; `version` command.
+  - Conventions from docsync/envdiff/cix/repolint.
+- [ ] **MCP server** (`cmd/mcp`, same module, stdio). Tools mirror the
+  CLI: `list_projects create_service deploy get_logs set_variable
+  restart rollback create_database trigger_cron deploy_template exec
+  get_metrics security_scan …`. Auth: `CONTAINR_URL` + `CONTAINR_TOKEN`
+  or shared `~/.containr/config`. `containr mcp serve` + install snippets
+  for Devin/Claude/Cursor.
+- [ ] **API agent ergonomics.** `Idempotency-Key` on mutating endpoints;
+  uniform `{error, code, details}` envelope (audit bare-string returns);
+  cursor pagination on growing lists; name-or-id lookup everywhere.
+- [ ] **Agent docs.** Root `AGENTS.md`, `/docs` serves `llms.txt` +
+  `llms-full.txt`, MCP tool list generated from OpenAPI where possible.
+
+**Gate**: an agent holding only a PAT registers a project, deploys from a
+git repo, streams logs, sets vars, rolls back — zero UI. Same via CLI
+for a human.
+
+---
+
+## 3. Phase B — Deployment pipeline maturity
+
+dflow's one real architectural edge. Containr deploys are fire-and-forget
+goroutines; this fixes correctness *and* unlocks the ops surface.
+
+- [ ] **Deploy job queue.** Lightweight internal queue (Postgres
+  `deployment_queue` table + single worker loop — no new infra; Redis
+  exists but don't add BullMQ-style sprawl). Serialize per service;
+  optional per-node serialization. Status gains `queued`.
+- [ ] **Cancel deployments.** Route `POST /deployments/:id/cancel` →
+  engine `CancelDeployment` (exists, unrouted). Queued jobs cancel
+  cleanly; building jobs cancel context where safe.
+- [ ] **Cache modes.** `?cache=full` on redeploy → rebuild from existing
+  image layers vs clean pull/checkout. Webhook-triggered deploys default
+  to no-cache (dflow convention — prevents stale-layer surprises).
+- [ ] **Queued→building→running status + persisted step logs.** Already
+  mostly there; add `queued`, stream over existing `/ws`.
+- [ ] **Env var apply → redeploy.** Saving variables asks "restart to
+  apply?" (or `--apply` flag in CLI); backend triggers redeploy honoring
+  running state. Currently rows update but containers keep stale env —
+  silent-drift bug class, not just UX.
+- [ ] **Encrypt variables at rest.** `environment_variables.value` is
+  plaintext today (verified). Envelope-encrypt `is_secret` values
+  (AES-GCM, key from `JWT_SECRET` or dedicated `SECRETS_KEY`); dflow
+  encrypts every var, we encrypt secrets — migration = transparent
+  encrypt-on-write + decrypt-on-read.
+
+## 4. Phase C — Service configuration parity
+
+- [ ] **Volumes on services** — `services.volumes` jsonb
+  `[{volume|host_path, container_path, read_only}]`; engine accepts
+  `VolumeMounts` already. Volumes tab + orphan-volume listing
+  (`docker volume ls -f dangling`) with delete/attach.
+- [ ] **Multi-domain** — `service_domains(service_id, domain, is_default,
+  cert_type, status, last_checked_at)`. Domains tab: add/remove/
+  set-default/DNS-status badge. Keep `services.domain` as derived
+  default for compat.
+- [ ] **DNS preflight** — `GET /services/:id/domains/check` resolves and
+  compares vs edge IP → `ok|wrong-target|pending`.
+- [ ] **Maintenance mode + basic-auth gate** — Traefik middleware labels;
+  no proxy code needed.
+- [ ] **Routing overrides** — `services.traefik_labels` jsonb with
+  allowlist validation (middlewares, headers, redirects).
+- [ ] **Builder override** — `services.builder` enum
+  `auto|railpack|nixpacks|dockerfile|static`; fix hardcoded `nixpacks`
+  at `deployments.go:406`. Static builder = build + nginx sidecar.
+- [ ] **Resource limits + reserves** — adopt dflow-style explicit units
+  (`"0.5"`/`"512m"`); add `cpu_reserve`/`memory_reserve` →
+  `NanoCPUs`/`Memory`/`MemoryReservation`; instance defaults in
+  `app_settings`.
+- [ ] **Pre-deploy capacity check** — sum reservations vs node capacity
+  (agent telemetry); warn or block per policy flag.
+- [ ] **Service clone** + **move between projects**.
+- [ ] **Private registries** — `registry_credentials`
+  (server/username/secret, encrypted); `docker login` at pull time;
+  per-service registry link.
+- [ ] **Service detail tabs to match dflow's surface**: add Domains,
+  Volumes, Proxy/access-controls, Backups sections to the service page
+  (Cron/Previews/Scaling already exist).
+
+## 5. Phase D — Data layer maturity
+
+- [ ] **External DB registration** — `provider='managed'|'external'`,
+  conn fields, `POST /databases/test-connection`, same bind flow.
+- [ ] **Public DB port expose** — opt-in Traefik TCP route or host port;
+  default stays loopback.
+- [ ] **Offsite backups** — `backup_targets` (S3-compatible:
+  endpoint/bucket/keys encrypted); per-DB destination; ship archives via
+  utility container + s3 SDK; restore-from-S3.
+- [ ] **Cross-node DB migration** — `POST /databases/:id/migrate
+  {target_node_id}`: dump → stream transfer → provision → restore →
+  re-point bound services. Needs ≥2 nodes or local↔node.
+- [ ] **Backup import** — upload archive → new restore point (pairs with
+  migrators).
+
+## 6. Phase E — Templates v2 (absorbs compose-catalog spec + dflow graph templates)
+
+- [ ] **Template = ordered service graph** — `services[]` entries with
+  `{name, type, source, build, ports, volumes, variables, depends_on}`;
+  deploy order from dependency refs.
+- [ ] **Expressions** — `${{service.KEY}}` (already the runtime ref
+  syntax — reuse), `{{secret(len,"charset")}}`, `{{random:int}}`,
+  `{{VAR:-default}}`; `POST /templates/:id/plan` returns the resolved
+  graph without deploying (agents love this).
+- [ ] **Compose import** — `POST /templates/import/compose` (compose-go
+  parser → graph), `POST /templates/import/github` (compose file from
+  repo path). Implements `docs/superpowers/specs/2026-04-14-*`.
+- [ ] **Graph deploy pipeline** — creates project + services in
+  dependency order, resolves vars, triggers builds; defined
+  partial-failure policy.
+- [ ] **Visual template builder** — canvas mode: drop git/image/db
+  nodes, wire refs, set order, save/publish. Reuses existing canvas
+  primitives (dflow's `templates/compose` is exactly this).
+- [ ] **Catalog growth** — convert `templates/*.md` (20 app guides) into
+  real graph templates; screenshots + badges per the old spec.
+
+**Gate**: paste n8n compose → 3-service project with generated secrets
+and wired refs, one click or one CLI call.
+
+## 7. Phase F — Nodes & infrastructure
+
+- [ ] **SSH bootstrap install** — `containr nodes add --ssh user@host` +
+  UI flow: one SSH session runs agent install (binary + systemd +
+  enroll token), SSH closed forever. dflow onboarding UX, agent-model
+  security.
+- [ ] **Mesh awareness** — agent detects + reports tailscale0/netbird0/
+  zt0 IPs at register; backend stores + displays. Optional
+  `TAILSCALE_AUTH_KEY` enroll-at-provision later.
+- [ ] **Node detail page** — placed services, telemetry history, agent
+  version + self-upgrade, **drain/cordon**, remove.
+- [ ] **Node housekeeping** — agent `prune` command (docker system
+  prune, bounded) + optional schedule (`setServerAutoCleanup` analog —
+  disk pressure is the #1 small-VPS killer).
+- [ ] **Multi-node scheduling** — wire `scheduling_rules` (table exists):
+  pin/spread/affinity, prefer-lowest-usage; deploy engine routes to
+  `services.node_id` via agent commands (agent already runs containers).
+- [ ] **Build-on-node** — agents currently `docker run` prebuilt images
+  only; extend agent to pull registry images (build stays central) —
+  private registry creds flow through Phase C.
+- [ ] **IaC provisioning (ROADMAP §2.2, unchanged)** — Terraform/
+  OpenTofu, `infra_connections`, Proxmox + AWS modules, + Hetzner module
+  (dflow's second provider), cloud-init → auto-enroll.
+- [ ] **Global/wildcard domain per node** (`nodes.default_domain`) —
+  auto-domains for services on that node.
+
+## 8. Phase G — Platform polish
+
+- [ ] **Outbound webhooks** — `webhooks(url, secret, events[], headers,
+  enabled)` + `webhook_deliveries` log; HMAC `X-Containr-Signature`;
+  events: `deployment.* service.* database.* security.* agent.*`.
+- [ ] **Activity feed** — enrich `audit_logs` with `severity category
+  icon label metadata`; per-project timeline + global feed; TTL job.
+- [ ] **Banners** — `banners(title, body, level, scope, active)`; top
+  bar in UI + surfaced to CLI on `auth status`.
+- [ ] **White-label** — `app_settings.branding`: product name, logos,
+  favicon, accent override, docs/support links. `GET /api/v1/branding`
+  public + cached. **Containr/Vertice remains default;** this is for
+  self-hosters who rebrand their instance.
+- [ ] **Operations page** — unified view of running/queued/failed jobs
+  (deploys, builds, backups, cron, scans, agent commands) — dflow's
+  bubble panel as a real page; flush/stuck-job tools once queue exists.
+- [ ] **Interactive terminal** — upgrade exec to WS PTY (`xterm.js`);
+  `nodes/:id/terminal` via agent (dflow's wetty without the extra
+  container).
+- [ ] **Team invites** — `POST /invites` → signed link → join flow;
+  replaces manual-only user creation.
+- [ ] **Impersonate user** — admin support tool, audit-logged.
+- [ ] **GitHub App manifest flow** — `POST /git/github-app/manifest`
+  returns manifest + redirect; instance self-provisions its App
+  (dflow pattern — removes the env-setup barrier).
+- [ ] **fork-sync trigger** — accept repo-sync events on webhook
+  receiver.
+- [ ] **Dynamic filter framework** — port dflow's `filter.utils`
+  pattern (typed filter configs → URL state → client filtering) for
+  audit logs, deployments, notifications tables.
+- [ ] **Notification producers round-out** — backups, agent offline,
+  upgrade available, security findings.
+- [ ] **Custom roles/RBAC** — defer until demanded; PAT scopes designed
+  forward-compatible now.
+
+## 9. Phase H — Migration & ecosystem
+
+- [ ] **Railway migrator** — `containr migrate railway --token --project
+  --dry-run`: Railway GraphQL → services/vars/domains → recreate via
+  PAT. Landing-page pitch: "off Railway in one command."
+- [ ] **dflow importer** — `containr migrate dflow` via Payload REST
+  export or Mongo dump: servers→nodes, projects, services(+vars/
+  domains/volumes), databases, templates. Onboards dflow refugees.
+- [ ] **`containr import compose`** — Phase-E parser straight to a
+  project (no template intermediate).
+- [ ] **Template registry** — optional `CONTAINR_TEMPLATE_REGISTRY_URL`
+  fetch/publish; bundled catalog works offline (dflow's self-contained
+  principle — nothing phones home).
+
+## 10. Carried-over debt (ROADMAP.md items still open)
+
+- [ ] **Gateway traffic path** — mount `/g/:slug/*` → key → rate-limit →
+  `gateway/proxy.go` → metrics. Management API exists; proxy unwired.
+- [ ] **Preview environments runtime** — real build+deploy+subdomain+TTL
+  (bookkeeping only now). Pairs with clone + multi-domain.
+- [ ] **Environments** — env switcher, per-env vars/domains, promote.
+- [ ] **sqlc port** — ~140 raw `database/sql` sites → `sqlc/queries/`.
+- [ ] **Tests** — canvas/wizard frontend coverage, deployment/template/
+  webhook table tests, Playwright e2e golden path.
+- [ ] **First-run wizard** — admin → GitHub App (manifest flow from
+  Phase G makes this self-serve) → Tunnel → done.
+- [ ] **Self-upgrade loop verification**, docs freshness, v1.0.0 release.
+
+## 11. Deliberately not imported from dflow
+
+- Dokku-over-SSH runtime — our Docker API + agent model is strictly
+  better.
+- Per-server-named BullMQ queues — our queue stays a single ordered
+  table.
+- Payload/Mongo/Next.js stack, multi-tenant plugin machinery.
+- Ansible playbook runner — arbitrary code-on-node is a security
+  surface; SSH bootstrap + agent commands cover the real needs.
+  Revisit only with a concrete demand.
+- `isAdmin`-only collection access — our owner/admin/public model is
+  richer already.
+- Netdata install flow — our agent telemetry + optional Beszel covers
+  monitoring; don't ship a third dashboard stack.
+
+## 12. Recommended order
+
+```
+A (agent surface) ──► B (deploy pipeline) ──► C/D (service+data depth)
+      │                                              │
+      ▼                                              ▼
+ CLI/MCP dogfood every feature              E (templates v2)
+                                                    │
+                                             F (nodes/infra) ──► H (migrators)
+                                                    │
+                                             G (polish) + debt
+```
+
+Why this order: PATs+CLI+MCP first = every later feature ships
+user-and-agent-visible at once. The queue (B) is prerequisite for
+the ops surface and correct concurrency. Templates v2 needs volumes,
+multi-domain, clone. Migrators last — they're launch marketing, not
+foundation.
+
+## 13. Design rules
+
+- **API is the contract.** UI, CLI, MCP are thin clients over
+  `openapi.yaml`. A feature isn't done until all three reach it.
+- **Everything scriptable.** JSON in/out, `--json`/`--yes`, uniform
+  pagination, name-or-id addressing.
+- **Same powers, same position.** Nothing user-facing that isn't
+  API-facing; agent-visible state (tokens, jobs) visible to users too.
+- **dflow is a donor, not a dependency.** Port semantics, not code.
+- **Containr branding is the default**; white-label is instance config.
+- **Self-hosted first** — no required external service, no phone-home.

app/backend/internal/api/routes.go

@@ -57,6 +57,33 @@ func SetupRoutes(router *gin.Engine, db *database.DB, redis *database.Redis, cfg
 		}
 	}
 	haManager := ha.NewHighAvailabilityManager(scheduler, metricsCollector)
+	if db != nil && db.DB != nil {
+		haManager.WithPersistence(db.DB)
+		if err := haManager.LoadHealthChecks(context.Background()); err != nil {
+			log.Printf("Failed to restore health checks: %v", err)
+		}
+	}
+	if dockerClient != nil && db != nil {
+		// Service-bound checks probe the service DNS name on the project
+		// network — the backend container joins it on demand. The old
+		// 127.0.0.1 fallback stays for node-bound checks.
+		docker := dockerClient
+		haManager.WithServiceResolver(func(ctx context.Context, serviceID string) (string, int, error) {
+			var name, projectID string
+			var port int
+			err := db.QueryRowContext(ctx,
+				`SELECT name, project_id::text, COALESCE(port, 0) FROM services WHERE id = $1`,
+				serviceID,
+			).Scan(&name, &projectID, &port)
+			if err != nil {
+				return "", 0, err
+			}
+			if err := docker.ConnectSelfToNetwork(ctx, deployment.ProjectNetworkName(projectID)); err != nil {
+				return "", 0, err
+			}
+			return name, port, nil
+		})
+	}
 	haAPIManager := NewHAManager(haManager)
 
 	// Initialize scaling handler
@@ -233,6 +260,11 @@ func SetupRoutes(router *gin.Engine, db *database.DB, redis *database.Redis, cfg
 			authed.GET("/user/profile", handleGetProfile)
 			authed.PUT("/user/profile", handleUpdateProfile)
 
+			// Personal access tokens for CLI/MCP/agent auth.
+			authed.GET("/user/tokens", handleListUserTokens)
+			authed.POST("/user/tokens", handleCreateUserToken)
+			authed.DELETE("/user/tokens/:id", handleDeleteUserToken)
+
 			authed.GET("/notifications", handleListNotifications)
 			authed.POST("/notifications/:id/read", handleMarkNotificationRead)
 			authed.POST("/notifications/read-all", handleMarkAllNotificationsRead)

app/backend/internal/api/user_tokens.go

@@ -0,0 +1,194 @@
+package api
+
+import (
+	"crypto/rand"
+	"crypto/sha256"
+	"database/sql"
+	"encoding/hex"
+	"net/http"
+	"strings"
+	"time"
+
+	"containr/internal/database"
+	sqlcdb "containr/internal/database/sqlcdb"
+
+	"github.com/gin-gonic/gin"
+	"github.com/google/uuid"
+)
+
+// Personal access tokens let the CLI, MCP server, and third-party agents
+// authenticate without a browser session. The raw token (cnp_…) is
+// returned exactly once at creation; only its sha256 hash persists.
+
+var userTokenScopes = map[string]bool{"read": true, "write": true, "admin": true}
+
+func generateUserToken() (token string, prefix string, hash string, err error) {
+	buf := make([]byte, 24)
+	if _, err = rand.Read(buf); err != nil {
+		return "", "", "", err
+	}
+	token = "cnp_" + hex.EncodeToString(buf)
+	prefix = token[:12] // cnp_ + 8 hex chars, enough to identify in a list
+	sum := sha256.Sum256([]byte(token))
+	return token, prefix, hex.EncodeToString(sum[:]), nil
+}
+
+func userTokenJSON(t sqlcdb.UserToken) gin.H {
+	row := gin.H{
+		"id":         t.ID.String(),
+		"name":       t.Name,
+		"key_prefix": t.KeyPrefix,
+		"scope":      t.Scope,
+	}
+	if t.CreatedAt.Valid {
+		row["created_at"] = t.CreatedAt.Time
+	}
+	if t.LastUsedAt.Valid {
+		row["last_used_at"] = t.LastUsedAt.Time
+	}
+	if t.ExpiresAt.Valid {
+		row["expires_at"] = t.ExpiresAt.Time
+	}
+	return row
+}
+
+func handleCreateUserToken(c *gin.Context) {
+	userID, ok := requireAuthenticatedUserID(c)
+	if !ok {
+		return
+	}
+	db := c.MustGet("db").(*database.DB)
+	queries := sqlcdb.New(db.DB)
+
+	var req struct {
+		Name          string `json:"name"`
+		Scope         string `json:"scope"`
+		ExpiresInDays int    `json:"expires_in_days"`
+	}
+	if err := c.ShouldBindJSON(&req); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+
+	req.Name = strings.TrimSpace(req.Name)
+	if req.Name == "" {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "name is required"})
+		return
+	}
+	if req.Scope == "" {
+		req.Scope = "write"
+	}
+	if !userTokenScopes[req.Scope] {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "scope must be one of: read, write, admin"})
+		return
+	}
+	// A token can never outrank its owner: non-admin users cannot mint
+	// admin-scope tokens even for themselves.
+	if req.Scope == "admin" && !c.GetBool("is_admin") {
+		c.JSON(http.StatusForbidden, gin.H{"error": "admin scope requires a platform admin account"})
+		return
+	}
+	if req.ExpiresInDays < 0 || req.ExpiresInDays > 3650 {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "expires_in_days must be between 0 (never) and 3650"})
+		return
+	}
+
+	token, prefix, hash, err := generateUserToken()
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "Failed to generate token"})
+		return
+	}
+
+	ownerUUID, err := uuid.Parse(userID)
+	if err != nil {
+		c.JSON(http.StatusUnauthorized, gin.H{"error": "Invalid user context"})
+		return
+	}
+
+	expiresAt := sql.NullTime{}
+	if req.ExpiresInDays > 0 {
+		expiresAt = sql.NullTime{Time: time.Now().Add(time.Duration(req.ExpiresInDays) * 24 * time.Hour), Valid: true}
+	}
+
+	row, err := queries.CreateUserToken(c.Request.Context(), sqlcdb.CreateUserTokenParams{
+		UserID:    ownerUUID,
+		Name:      req.Name,
+		KeyPrefix: prefix,
+		TokenHash: hash,
+		Scope:     req.Scope,
+		ExpiresAt: expiresAt,
+	})
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "Failed to store token"})
+		return
+	}
+
+	LogAudit(userID, "user_token", row.ID.String(), "create", map[string]interface{}{"name": req.Name, "scope": req.Scope})
+
+	payload := userTokenJSON(row)
+	payload["token"] = token
+	c.JSON(http.StatusCreated, payload)
+}
+
+func handleListUserTokens(c *gin.Context) {
+	userID, ok := requireAuthenticatedUserID(c)
+	if !ok {
+		return
+	}
+	db := c.MustGet("db").(*database.DB)
+	queries := sqlcdb.New(db.DB)
+
+	ownerUUID, err := uuid.Parse(userID)
+	if err != nil {
+		c.JSON(http.StatusUnauthorized, gin.H{"error": "Invalid user context"})
+		return
+	}
+
+	rows, err := queries.ListUserTokens(c.Request.Context(), ownerUUID)
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "Failed to fetch tokens"})
+		return
+	}
+
+	out := make([]gin.H, 0, len(rows))
+	for _, t := range rows {
+		out = append(out, userTokenJSON(t))
+	}
+	c.JSON(http.StatusOK, gin.H{"tokens": out})
+}
+
+func handleDeleteUserToken(c *gin.Context) {
+	userID, ok := requireAuthenticatedUserID(c)
+	if !ok {
+		return
+	}
+	db := c.MustGet("db").(*database.DB)
+	queries := sqlcdb.New(db.DB)
+
+	ownerUUID, err := uuid.Parse(userID)
+	if err != nil {
+		c.JSON(http.StatusUnauthorized, gin.H{"error": "Invalid user context"})
+		return
+	}
+	tokenUUID, err := uuid.Parse(c.Param("id"))
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "Invalid token ID"})
+		return
+	}
+
+	affected, err := queries.RevokeUserToken(c.Request.Context(), sqlcdb.RevokeUserTokenParams{
+		ID:     tokenUUID,
+		UserID: ownerUUID,
+	})
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "Failed to revoke token"})
+		return
+	}
+	if affected == 0 {
+		c.JSON(http.StatusNotFound, gin.H{"error": "Token not found"})
+		return
+	}
+
+	LogAudit(userID, "user_token", tokenUUID.String(), "revoke", nil)
+	c.JSON(http.StatusOK, gin.H{"message": "Token revoked"})
+}

app/backend/internal/database/sqlcdb/models.go

@@ -195,6 +195,12 @@ type Service struct {
 	BuildPath       sql.NullString `json:"build_path"`
 	Cpu             sql.NullString `json:"cpu"`
 	Memory          sql.NullString `json:"memory"`
+	Replicas        int32          `json:"replicas"`
+	Port            int32          `json:"port"`
+	Domain          string         `json:"domain"`
+	HealthcheckPath string         `json:"healthcheck_path"`
+	RestartPolicy   string         `json:"restart_policy"`
+	PublishedPort   int32          `json:"published_port"`
 }
 
 type ServiceTemplate struct {
@@ -211,3 +217,16 @@ type ServiceTemplate struct {
 	CreatedAt   sql.NullTime          `json:"created_at"`
 	UpdatedAt   sql.NullTime          `json:"updated_at"`
 }
+
+type UserToken struct {
+	ID         uuid.UUID    `json:"id"`
+	UserID     uuid.UUID    `json:"user_id"`
+	Name       string       `json:"name"`
+	KeyPrefix  string       `json:"key_prefix"`
+	TokenHash  string       `json:"token_hash"`
+	Scope      string       `json:"scope"`
+	ExpiresAt  sql.NullTime `json:"expires_at"`
+	LastUsedAt sql.NullTime `json:"last_used_at"`
+	RevokedAt  sql.NullTime `json:"revoked_at"`
+	CreatedAt  sql.NullTime `json:"created_at"`
+}

app/backend/internal/database/sqlcdb/querier.go

@@ -26,6 +26,7 @@ type Querier interface {
 	CreateProject(ctx context.Context, arg CreateProjectParams) (Project, error)
 	CreateServiceFromTemplate(ctx context.Context, arg CreateServiceFromTemplateParams) error
 	CreateUserTemplate(ctx context.Context, arg CreateUserTemplateParams) error
+	CreateUserToken(ctx context.Context, arg CreateUserTokenParams) (UserToken, error)
 	DatabaseServiceExistsByIDAndUser(ctx context.Context, arg DatabaseServiceExistsByIDAndUserParams) (bool, error)
 	DeleteAgent(ctx context.Context, id string) error
 	DeleteDatabaseServiceByIDAndUser(ctx context.Context, arg DeleteDatabaseServiceByIDAndUserParams) error
@@ -46,6 +47,7 @@ type Querier interface {
 	GetProjectOwnerID(ctx context.Context, id uuid.UUID) (uuid.UUID, error)
 	GetProjectRoleForUser(ctx context.Context, arg GetProjectRoleForUserParams) (string, error)
 	GetServiceTemplateByID(ctx context.Context, id string) (ServiceTemplate, error)
+	GetUserTokenByHash(ctx context.Context, tokenHash string) (UserToken, error)
 	InsertAgentHeartbeat(ctx context.Context, arg InsertAgentHeartbeatParams) error
 	InsertProjectEnvironment(ctx context.Context, arg InsertProjectEnvironmentParams) error
 	ListAgentAuthTokens(ctx context.Context) ([]AgentAuthToken, error)
@@ -66,9 +68,11 @@ type Querier interface {
 	ListProjectsWithStatsByUser(ctx context.Context, arg ListProjectsWithStatsByUserParams) ([]ListProjectsWithStatsByUserRow, error)
 	ListServiceTemplatesByCategoryForUser(ctx context.Context, arg ListServiceTemplatesByCategoryForUserParams) ([]ServiceTemplate, error)
 	ListServiceTemplatesForUser(ctx context.Context, ownerID uuid.NullUUID) ([]ServiceTemplate, error)
+	ListUserTokens(ctx context.Context, userID uuid.UUID) ([]UserToken, error)
 	MarkAllNotificationsReadByUser(ctx context.Context, arg MarkAllNotificationsReadByUserParams) error
 	MarkNotificationReadByIDAndUser(ctx context.Context, arg MarkNotificationReadByIDAndUserParams) error
 	RevokeAgentAuthToken(ctx context.Context, id uuid.UUID) (AgentAuthToken, error)
+	RevokeUserToken(ctx context.Context, arg RevokeUserTokenParams) (int64, error)
 	SetDatabaseBackupScheduleByIDAndUser(ctx context.Context, arg SetDatabaseBackupScheduleByIDAndUserParams) error
 	SetDatabaseBackupStatusByID(ctx context.Context, arg SetDatabaseBackupStatusByIDParams) error
 	SetDatabaseNextBackupAt(ctx context.Context, arg SetDatabaseNextBackupAtParams) error
@@ -77,6 +81,7 @@ type Querier interface {
 	SetDatabaseServiceStatusByIDAndUser(ctx context.Context, arg SetDatabaseServiceStatusByIDAndUserParams) error
 	SetProjectApproved(ctx context.Context, arg SetProjectApprovedParams) (int64, error)
 	TouchAgentAuthToken(ctx context.Context, tokenHash string) error
+	TouchUserToken(ctx context.Context, id uuid.UUID) error
 	UpdateAgent(ctx context.Context, arg UpdateAgentParams) (NodeAgent, error)
 	UpdateAgentHeartbeat(ctx context.Context, arg UpdateAgentHeartbeatParams) error
 	UpdateContainerStatus(ctx context.Context, arg UpdateContainerStatusParams) error

app/backend/internal/database/sqlcdb/user_tokens.sql.go

@@ -0,0 +1,147 @@
+// Code generated by sqlc. DO NOT EDIT.
+// versions:
+//   sqlc v1.31.1
+// source: user_tokens.sql
+
+package sqlcdb
+
+import (
+	"context"
+	"database/sql"
+
+	"github.com/google/uuid"
+)
+
+const createUserToken = `-- name: CreateUserToken :one
+INSERT INTO user_tokens (user_id, name, key_prefix, token_hash, scope, expires_at)
+VALUES ($1, $2, $3, $4, $5, $6)
+RETURNING id, user_id, name, key_prefix, token_hash, scope, expires_at, last_used_at, revoked_at, created_at
+`
+
+type CreateUserTokenParams struct {
+	UserID    uuid.UUID    `json:"user_id"`
+	Name      string       `json:"name"`
+	KeyPrefix string       `json:"key_prefix"`
+	TokenHash string       `json:"token_hash"`
+	Scope     string       `json:"scope"`
+	ExpiresAt sql.NullTime `json:"expires_at"`
+}
+
+func (q *Queries) CreateUserToken(ctx context.Context, arg CreateUserTokenParams) (UserToken, error) {
+	row := q.db.QueryRowContext(ctx, createUserToken,
+		arg.UserID,
+		arg.Name,
+		arg.KeyPrefix,
+		arg.TokenHash,
+		arg.Scope,
+		arg.ExpiresAt,
+	)
+	var i UserToken
+	err := row.Scan(
+		&i.ID,
+		&i.UserID,
+		&i.Name,
+		&i.KeyPrefix,
+		&i.TokenHash,
+		&i.Scope,
+		&i.ExpiresAt,
+		&i.LastUsedAt,
+		&i.RevokedAt,
+		&i.CreatedAt,
+	)
+	return i, err
+}
+
+const getUserTokenByHash = `-- name: GetUserTokenByHash :one
+SELECT id, user_id, name, key_prefix, token_hash, scope, expires_at, last_used_at, revoked_at, created_at FROM user_tokens
+WHERE token_hash = $1
+  AND revoked_at IS NULL
+  AND (expires_at IS NULL OR expires_at > NOW())
+LIMIT 1
+`
+
+func (q *Queries) GetUserTokenByHash(ctx context.Context, tokenHash string) (UserToken, error) {
+	row := q.db.QueryRowContext(ctx, getUserTokenByHash, tokenHash)
+	var i UserToken
+	err := row.Scan(
+		&i.ID,
+		&i.UserID,
+		&i.Name,
+		&i.KeyPrefix,
+		&i.TokenHash,
+		&i.Scope,
+		&i.ExpiresAt,
+		&i.LastUsedAt,
+		&i.RevokedAt,
+		&i.CreatedAt,
+	)
+	return i, err
+}
+
+const listUserTokens = `-- name: ListUserTokens :many
+SELECT id, user_id, name, key_prefix, token_hash, scope, expires_at, last_used_at, revoked_at, created_at FROM user_tokens
+WHERE user_id = $1 AND revoked_at IS NULL
+ORDER BY created_at DESC
+`
+
+func (q *Queries) ListUserTokens(ctx context.Context, userID uuid.UUID) ([]UserToken, error) {
+	rows, err := q.db.QueryContext(ctx, listUserTokens, userID)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+	items := []UserToken{}
+	for rows.Next() {
+		var i UserToken
+		if err := rows.Scan(
+			&i.ID,
+			&i.UserID,
+			&i.Name,
+			&i.KeyPrefix,
+			&i.TokenHash,
+			&i.Scope,
+			&i.ExpiresAt,
+			&i.LastUsedAt,
+			&i.RevokedAt,
+			&i.CreatedAt,
+		); err != nil {
+			return nil, err
+		}
+		items = append(items, i)
+	}
+	if err := rows.Close(); err != nil {
+		return nil, err
+	}
+	if err := rows.Err(); err != nil {
+		return nil, err
+	}
+	return items, nil
+}
+
+const revokeUserToken = `-- name: RevokeUserToken :execrows
+UPDATE user_tokens SET revoked_at = NOW()
+WHERE id = $1 AND user_id = $2 AND revoked_at IS NULL
+`
+
+type RevokeUserTokenParams struct {
+	ID     uuid.UUID `json:"id"`
+	UserID uuid.UUID `json:"user_id"`
+}
+
+func (q *Queries) RevokeUserToken(ctx context.Context, arg RevokeUserTokenParams) (int64, error) {
+	result, err := q.db.ExecContext(ctx, revokeUserToken, arg.ID, arg.UserID)
+	if err != nil {
+		return 0, err
+	}
+	return result.RowsAffected()
+}
+
+const touchUserToken = `-- name: TouchUserToken :exec
+UPDATE user_tokens SET last_used_at = NOW()
+WHERE id = $1 AND revoked_at IS NULL
+`
+
+func (q *Queries) TouchUserToken(ctx context.Context, id uuid.UUID) error {
+	_, err := q.db.ExecContext(ctx, touchUserToken, id)
+	return err
+}

app/backend/internal/deployment/runtime.go

@@ -67,7 +67,9 @@ type RuntimeState struct {
 	Status     string             `json:"status"`           // running, degraded, stopped
 }
 
-func projectNetworkName(projectID string) string {
+// ProjectNetworkName is the deterministic docker network every service in a
+// project joins — containr-proj-<first 12 hex chars of the project uuid>.
+func ProjectNetworkName(projectID string) string {
 	short := strings.ReplaceAll(projectID, "-", "")
 	if len(short) > 12 {
 		short = short[:12]
@@ -82,7 +84,7 @@ func serviceNamePrefix(serviceID string) string {
 // EnsureProjectNetwork creates the per-project bridge network if missing.
 // Containers on it resolve each other by service name (private networking).
 func (de *DeploymentEngine) EnsureProjectNetwork(ctx context.Context, projectID string) (string, error) {
-	name := projectNetworkName(projectID)
+	name := ProjectNetworkName(projectID)
 
 	networks, err := de.dockerClient.ListNetworks(ctx)
 	if err != nil {
@@ -364,7 +366,7 @@ func (de *DeploymentEngine) ProbeServiceHealth(ctx context.Context, state *Runti
 	}
 
 	probeURL := ""
-	if networkName := projectNetworkName(projectID); networkName != "" {
+	if networkName := ProjectNetworkName(projectID); networkName != "" {
 		if err := de.dockerClient.ConnectSelfToNetwork(ctx, networkName); err == nil {
 			probeURL = fmt.Sprintf("http://%s:%d%s", serviceName, port, healthPath)
 		}
@@ -456,7 +458,7 @@ func (de *DeploymentEngine) RemoveProjectContainers(ctx context.Context, project
 
 // RemoveProjectNetwork removes the per-project bridge network if present.
 func (de *DeploymentEngine) RemoveProjectNetwork(ctx context.Context, projectID string) error {
-	return de.dockerClient.RemoveNetwork(ctx, projectNetworkName(projectID))
+	return de.dockerClient.RemoveNetwork(ctx, ProjectNetworkName(projectID))
 }
 
 func replicaIndex(c container.Summary) int {

app/backend/internal/ha/manager.go

@@ -3,6 +3,7 @@ package ha
 import (
 	"bytes"
 	"context"
+	"database/sql"
 	"encoding/json"
 	"fmt"
 	"log"
@@ -80,6 +81,8 @@ type HealthChecker struct {
 	results          map[string]*HealthCheckResult
 	mu               sync.RWMutex
 	checkInterval    time.Duration
+	serviceResolver  func(ctx context.Context, serviceID string) (host string, port int, err error)
+	db               *sql.DB
 }
 
 // HealthCheck represents a health check configuration
@@ -596,7 +599,11 @@ func (hc *HealthChecker) performHealthChecks(ctx context.Context) error {
 		result := hc.performHealthCheck(ctx, check)
 		hc.mu.Lock()
 		hc.results[check.ID] = result
+		check.Status = result.Status
+		check.LastCheck = result.Timestamp
 		hc.mu.Unlock()
+		// Persist last-known state so a restart resumes with honest history.
+		hc.persistCheck(check)
 	}
 
 	return nil
@@ -617,12 +624,20 @@ func (hc *HealthChecker) performHealthCheck(ctx context.Context, check *HealthCh
 	checkCtx, cancel := context.WithTimeout(ctx, timeout)
 	defer cancel()
 
-	var err error
+	host, port, err := hc.resolveTarget(ctx, check)
+	if err != nil {
+		result.Status = HealthStatusUnhealthy
+		result.Message = err.Error()
+		result.ErrorCode = "TARGET_UNRESOLVED"
+		result.Latency = time.Since(start)
+		return result
+	}
+
 	switch check.Type {
 	case HealthCheckTypeHTTP:
-		err = hc.performHTTPCheck(checkCtx, check)
+		err = hc.performHTTPCheck(checkCtx, check, host, port)
 	case HealthCheckTypeTCP:
-		err = hc.performTCPCheck(checkCtx, check)
+		err = hc.performTCPCheck(checkCtx, check, host, port)
 	case HealthCheckTypeCommand:
 		err = hc.performCommandCheck(checkCtx, check)
 	default:
@@ -642,13 +657,11 @@ func (hc *HealthChecker) performHealthCheck(ctx context.Context, check *HealthCh
 	return result
 }
 
-func (hc *HealthChecker) performHTTPCheck(ctx context.Context, check *HealthCheck) error {
+func (hc *HealthChecker) performHTTPCheck(ctx context.Context, check *HealthCheck, host string, port int) error {
-	host := hc.resolveHealthCheckHost(check)
 	protocol := strings.ToLower(strings.TrimSpace(check.Config.Protocol))
 	if protocol == "" {
 		protocol = "http"
 	}
-	port := check.Config.Port
 	if port <= 0 {
 		if protocol == "https" {
 			port = 443
@@ -684,9 +697,7 @@ func (hc *HealthChecker) performHTTPCheck(ctx context.Context, check *HealthChec
 	return nil
 }
 
-func (hc *HealthChecker) performTCPCheck(ctx context.Context, check *HealthCheck) error {
+func (hc *HealthChecker) performTCPCheck(ctx context.Context, check *HealthCheck, host string, port int) error {
-	host := hc.resolveHealthCheckHost(check)
-	port := check.Config.Port
 	if port <= 0 {
 		port = 80
 	}
@@ -723,6 +734,27 @@ func (hc *HealthChecker) performCommandCheck(ctx context.Context, check *HealthC
 	return nil
 }
 
+// WithServiceResolver lets the API layer plug in service-aware targeting: a
+// ServiceID-bound check resolves to the service's DNS name on its project
+// network instead of the backend container's loopback, where dialing
+// 127.0.0.1:<host port> can never succeed.
+func (m *HighAvailabilityManager) WithServiceResolver(fn func(ctx context.Context, serviceID string) (host string, port int, err error)) *HighAvailabilityManager {
+	m.healthChecker.serviceResolver = fn
+	return m
+}
+
+// WithPersistence attaches Postgres so configured checks survive restarts.
+// Call LoadHealthChecks once during startup to hydrate state.
+func (m *HighAvailabilityManager) WithPersistence(db *sql.DB) *HighAvailabilityManager {
+	m.healthChecker.db = db
+	return m
+}
+
+// LoadHealthChecks hydrates persisted health checks during startup.
+func (m *HighAvailabilityManager) LoadHealthChecks(ctx context.Context) error {
+	return m.healthChecker.LoadHealthChecks(ctx)
+}
+
 func (hc *HealthChecker) resolveHealthCheckHost(check *HealthCheck) string {
 	if strings.TrimSpace(check.NodeID) != "" && hc.scheduler != nil {
 		nodes := hc.scheduler.GetNodes()
@@ -739,19 +771,109 @@ func (hc *HealthChecker) resolveHealthCheckHost(check *HealthCheck) string {
 	return "127.0.0.1"
 }
 
+// resolveTarget picks the dial target for a check: service-bound checks go to
+// the service DNS name on the project network; node-bound and free-standing
+// checks keep the old host behavior. A service-bound resolver failure is
+// surfaced as an error — falling back to loopback would only produce a
+// misleading "service down" result.
+func (hc *HealthChecker) resolveTarget(ctx context.Context, check *HealthCheck) (string, int, error) {
+	port := check.Config.Port
+	if strings.TrimSpace(check.NodeID) == "" && check.ServiceID != "" && hc.serviceResolver != nil {
+		resolveCtx, cancel := context.WithTimeout(ctx, 10*time.Second)
+		defer cancel()
+		h, p, err := hc.serviceResolver(resolveCtx, check.ServiceID)
+		if err != nil {
+			return "", 0, fmt.Errorf("resolve service target: %w", err)
+		}
+		if h == "" {
+			return "", 0, fmt.Errorf("resolve service target: service %s not found", check.ServiceID)
+		}
+		if port <= 0 {
+			port = p
+		}
+		return h, port, nil
+	}
+	return hc.resolveHealthCheckHost(check), port, nil
+}
+
 // AddHealthCheck adds a new health check
 func (hc *HealthChecker) AddHealthCheck(check *HealthCheck) {
 	hc.mu.Lock()
-	defer hc.mu.Unlock()
 	hc.checks[check.ID] = check
+	hc.mu.Unlock()
+	hc.persistCheck(check)
 }
 
 // RemoveHealthCheck removes a health check
 func (hc *HealthChecker) RemoveHealthCheck(checkID string) {
 	hc.mu.Lock()
-	defer hc.mu.Unlock()
 	delete(hc.checks, checkID)
 	delete(hc.results, checkID)
+	hc.mu.Unlock()
+	if hc.db != nil {
+		_, _ = hc.db.ExecContext(context.Background(), `DELETE FROM ha_health_checks WHERE id = $1`, checkID)
+	}
+}
+
+// persistCheck upserts one check. Best effort — a write failure never blocks
+// the in-memory path, matching the scaling-policy persistence style.
+func (hc *HealthChecker) persistCheck(check *HealthCheck) {
+	if hc.db == nil {
+		return
+	}
+	cfg, _ := json.Marshal(check.Config)
+	var lastCheck interface{}
+	if !check.LastCheck.IsZero() {
+		lastCheck = check.LastCheck
+	}
+	_, _ = hc.db.ExecContext(context.Background(),
+		`INSERT INTO ha_health_checks (id, service_id, node_id, type, config, status, last_check, updated_at)
+		 VALUES ($1, $2, $3, $4, $5, $6, $7, now())
+		 ON CONFLICT (id) DO UPDATE
+		 SET service_id = EXCLUDED.service_id, node_id = EXCLUDED.node_id,
+		     type = EXCLUDED.type, config = EXCLUDED.config,
+		     status = EXCLUDED.status, last_check = EXCLUDED.last_check,
+		     updated_at = now()`,
+		check.ID, check.ServiceID, check.NodeID, string(check.Type), cfg, string(check.Status), lastCheck,
+	)
+}
+
+// LoadHealthChecks hydrates checks persisted across restarts.
+func (hc *HealthChecker) LoadHealthChecks(ctx context.Context) error {
+	if hc.db == nil {
+		return nil
+	}
+	rows, err := hc.db.QueryContext(ctx,
+		`SELECT id, service_id, node_id, type, config, status, last_check FROM ha_health_checks`)
+	if err != nil {
+		return err
+	}
+	defer rows.Close()
+
+	loaded := 0
+	for rows.Next() {
+		var check HealthCheck
+		var cfg []byte
+		var status, checkType string
+		var lastCheck sql.NullTime
+		if err := rows.Scan(&check.ID, &check.ServiceID, &check.NodeID, &checkType, &cfg, &status, &lastCheck); err != nil {
+			continue
+		}
+		check.Type = HealthCheckType(checkType)
+		check.Status = HealthStatus(status)
+		if lastCheck.Valid {
+			check.LastCheck = lastCheck.Time
+		}
+		_ = json.Unmarshal(cfg, &check.Config)
+		hc.mu.Lock()
+		hc.checks[check.ID] = &check
+		hc.mu.Unlock()
+		loaded++
+	}
+	if loaded > 0 {
+		log.Printf("Restored %d health checks", loaded)
+	}
+	return rows.Err()
 }
 
 // GetHealthCheck returns one configured health check.

app/backend/internal/middleware/middleware.go

@@ -2,7 +2,10 @@ package middleware
 
 import (
 	"containr/internal/database"
+	sqlcdb "containr/internal/database/sqlcdb"
+	"crypto/sha256"
 	"database/sql"
+	"encoding/hex"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -75,6 +78,14 @@ func RequestID() gin.HandlerFunc {
 func resolveRequestUser(c *gin.Context, jwtSecret string, sessionVerifier *betterAuthSessionVerifier) (userID, email, tokenErr string, hasToken, ok bool) {
 	tokenString, tokenErr, hasToken := extractJWTToken(c)
 	if tokenString != "" {
+		// Personal access tokens (cnp_…) are not JWTs — they resolve
+		// against the user_tokens table by sha256 hash.
+		if strings.HasPrefix(tokenString, patPrefix) {
+			if id, mail, verified := resolvePersonalAccessToken(c, tokenString); verified {
+				return id, mail, "", hasToken, true
+			}
+			return "", "", "Invalid token", hasToken, false
+		}
 		if claims, valid := validateJWTClaims(tokenString, jwtSecret); valid {
 			userIDClaim, exists := claims["user_id"]
 			if exists {
@@ -101,12 +112,67 @@ func resolveRequestUser(c *gin.Context, jwtSecret string, sessionVerifier *bette
 	return "", "", tokenErr, hasToken, false
 }
 
+const patPrefix = "cnp_"
+
+// resolvePersonalAccessToken authenticates a `cnp_` bearer token against the
+// user_tokens table (sha256 hash match, not revoked, not expired). On success
+// it stamps last_used_at, records the token scope on the context, and
+// returns the owner's id and email.
+func resolvePersonalAccessToken(c *gin.Context, token string) (string, string, bool) {
+	dbValue, exists := c.Get("db")
+	if !exists {
+		return "", "", false
+	}
+	db, ok := dbValue.(*database.DB)
+	if !ok || db == nil || db.DB == nil {
+		return "", "", false
+	}
+
+	sum := sha256.Sum256([]byte(token))
+	queries := sqlcdb.New(db.DB)
+	row, err := queries.GetUserTokenByHash(c.Request.Context(), hex.EncodeToString(sum[:]))
+	if err != nil {
+		return "", "", false
+	}
+
+	var email string
+	if err := db.QueryRowContext(c.Request.Context(),
+		`SELECT email FROM users WHERE id = $1`, row.UserID).Scan(&email); err != nil {
+		return "", "", false
+	}
+
+	_ = queries.TouchUserToken(c.Request.Context(), row.ID)
+	c.Set("pat_scope", row.Scope)
+	return row.UserID.String(), email, true
+}
+
+// patScopeAllows reports whether a PAT-authenticated request may proceed.
+// read scope permits only safe methods; write and admin are unrestricted at
+// this layer (admin routes still require the admin scope AND a platform
+// admin account — enforced via is_admin in setAuthenticatedContext).
+func patScopeAllows(c *gin.Context) bool {
+	if c.GetString("pat_scope") != "read" {
+		return true
+	}
+	switch c.Request.Method {
+	case http.MethodGet, http.MethodHead, http.MethodOptions:
+		return true
+	}
+	return false
+}
+
 // setAuthenticatedContext stores the resolved identity plus the admin flag on
 // the gin context so handlers and RequireAdmin can rely on it.
 func setAuthenticatedContext(c *gin.Context, userID, email string) {
 	c.Set("user_id", userID)
 	c.Set("email", email)
-	c.Set("is_admin", loadIsAdmin(c, userID))
+	isAdmin := loadIsAdmin(c, userID)
+	// A PAT can never outrank its scope: non-admin scopes cap the token at
+	// non-admin even when the owning user is a platform admin.
+	if scope := c.GetString("pat_scope"); scope != "" && scope != "admin" {
+		isAdmin = false
+	}
+	c.Set("is_admin", isAdmin)
 }
 
 func loadIsAdmin(c *gin.Context, userID string) bool {
@@ -133,6 +199,14 @@ func Auth(jwtSecret string) gin.HandlerFunc {
 		userID, email, tokenErr, hasToken, ok := resolveRequestUser(c, jwtSecret, sessionVerifier)
 		if ok {
 			setAuthenticatedContext(c, userID, email)
+			if !patScopeAllows(c) {
+				c.JSON(http.StatusForbidden, gin.H{
+					"error": "Token scope is read-only",
+					"code":  "SCOPE_INSUFFICIENT",
+				})
+				c.Abort()
+				return
+			}
 			c.Next()
 			return
 		}

app/backend/internal/middleware/middleware_test.go

@@ -131,6 +131,61 @@ func TestAuthStoresStringUserIDForValidClaims(t *testing.T) {
 	}
 }
 
+func TestAuthRejectsPATWithoutDatabase(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	router := gin.New()
+	router.Use(Auth("secret"))
+	router.GET("/test", func(c *gin.Context) {
+		c.String(http.StatusOK, "ok")
+	})
+
+	// A cnp_-prefixed token takes the PAT path, not JWT parsing — without a
+	// db handle it must fail closed with 401.
+	req := httptest.NewRequest(http.MethodGet, "/test", nil)
+	req.Header.Set("Authorization", "Bearer cnp_"+strings.Repeat("ab", 24))
+	rec := httptest.NewRecorder()
+	router.ServeHTTP(rec, req)
+
+	if rec.Code != http.StatusUnauthorized {
+		t.Fatalf("expected status %d, got %d", http.StatusUnauthorized, rec.Code)
+	}
+}
+
+func TestPATScopeAllows(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+
+	cases := []struct {
+		name   string
+		scope  string
+		method string
+		want   bool
+	}{
+		{"read allows GET", "read", http.MethodGet, true},
+		{"read allows HEAD", "read", http.MethodHead, true},
+		{"read allows OPTIONS", "read", http.MethodOptions, true},
+		{"read blocks POST", "read", http.MethodPost, false},
+		{"read blocks DELETE", "read", http.MethodDelete, false},
+		{"write allows POST", "write", http.MethodPost, true},
+		{"write allows DELETE", "write", http.MethodDelete, true},
+		{"admin allows PATCH", "admin", http.MethodPatch, true},
+		{"no scope (session/JWT) unrestricted", "", http.MethodPost, true},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			rec := httptest.NewRecorder()
+			c, _ := gin.CreateTestContext(rec)
+			c.Request = httptest.NewRequest(tc.method, "/x", nil)
+			if tc.scope != "" {
+				c.Set("pat_scope", tc.scope)
+			}
+			if got := patScopeAllows(c); got != tc.want {
+				t.Fatalf("patScopeAllows(scope=%q, method=%s) = %v, want %v", tc.scope, tc.method, got, tc.want)
+			}
+		})
+	}
+}
+
 func issueJWT(t *testing.T, secret string, claims jwt.MapClaims) string {
 	t.Helper()
 	token := jwt.NewWithClaims(jwt.SigningMethodHS256, claims)

app/backend/migrations_goose/20261004120000_ha_health_checks.sql

@@ -0,0 +1,18 @@
+-- +goose Up
+-- Persisted HA health checks. Previously checks lived only in memory, so a
+-- backend restart silently deleted every configured check while services kept
+-- running unmonitored.
+CREATE TABLE IF NOT EXISTS ha_health_checks (
+    id          uuid PRIMARY KEY,
+    service_id  uuid NOT NULL REFERENCES services(id) ON DELETE CASCADE,
+    node_id     varchar(255) NOT NULL DEFAULT '',
+    type        varchar(20) NOT NULL,
+    config      jsonb NOT NULL DEFAULT '{}'::jsonb,
+    status      varchar(20) NOT NULL DEFAULT 'unknown',
+    last_check  timestamptz,
+    created_at  timestamptz NOT NULL DEFAULT now(),
+    updated_at  timestamptz NOT NULL DEFAULT now()
+);
+
+-- +goose Down
+DROP TABLE IF EXISTS ha_health_checks;

app/backend/migrations_goose/20261008000000_user_tokens.sql

@@ -0,0 +1,25 @@
+-- +goose Up
+-- Personal access tokens for CLI/MCP/agent authentication.
+-- Only sha256 hashes are stored — raw tokens (cnp_…) are shown once at
+-- creation and never persisted. Scope gates what the token may do:
+-- read = GET/HEAD only, write = all non-admin routes, admin = everything
+-- (admin still requires the owning user to be a platform admin).
+CREATE TABLE IF NOT EXISTS public.user_tokens (
+    id uuid DEFAULT gen_random_uuid() NOT NULL PRIMARY KEY,
+    user_id uuid NOT NULL REFERENCES public.users(id) ON DELETE CASCADE,
+    name character varying(255) NOT NULL,
+    key_prefix character varying(16) NOT NULL,
+    token_hash text NOT NULL UNIQUE,
+    scope character varying(20) NOT NULL DEFAULT 'write'
+        CHECK (scope IN ('read', 'write', 'admin')),
+    expires_at timestamp with time zone,
+    last_used_at timestamp with time zone,
+    revoked_at timestamp with time zone,
+    created_at timestamp with time zone DEFAULT now() NOT NULL
+);
+
+CREATE INDEX IF NOT EXISTS idx_user_tokens_user_id ON public.user_tokens(user_id);
+CREATE INDEX IF NOT EXISTS idx_user_tokens_token_hash ON public.user_tokens(token_hash);
+
+-- +goose Down
+DROP TABLE IF EXISTS public.user_tokens;

app/backend/sqlc/queries/user_tokens.sql

@@ -0,0 +1,24 @@
+-- name: CreateUserToken :one
+INSERT INTO user_tokens (user_id, name, key_prefix, token_hash, scope, expires_at)
+VALUES ($1, $2, $3, $4, $5, $6)
+RETURNING *;
+
+-- name: ListUserTokens :many
+SELECT * FROM user_tokens
+WHERE user_id = $1 AND revoked_at IS NULL
+ORDER BY created_at DESC;
+
+-- name: GetUserTokenByHash :one
+SELECT * FROM user_tokens
+WHERE token_hash = $1
+  AND revoked_at IS NULL
+  AND (expires_at IS NULL OR expires_at > NOW())
+LIMIT 1;
+
+-- name: TouchUserToken :exec
+UPDATE user_tokens SET last_used_at = NOW()
+WHERE id = $1 AND revoked_at IS NULL;
+
+-- name: RevokeUserToken :execrows
+UPDATE user_tokens SET revoked_at = NOW()
+WHERE id = $1 AND user_id = $2 AND revoked_at IS NULL;

app/backend/sqlc/schema.sql

@@ -52,7 +52,13 @@ CREATE TABLE services (
     git_branch VARCHAR(100),
     build_path VARCHAR(500),
     cpu VARCHAR(50),
-    memory VARCHAR(50)
+    memory VARCHAR(50),
+    replicas INTEGER NOT NULL DEFAULT 1,
+    port INTEGER NOT NULL DEFAULT 0,
+    domain VARCHAR(255) NOT NULL DEFAULT '',
+    healthcheck_path VARCHAR(255) NOT NULL DEFAULT '',
+    restart_policy VARCHAR(50) NOT NULL DEFAULT 'unless-stopped',
+    published_port INTEGER NOT NULL DEFAULT 0
 );
 
 CREATE TABLE deployments (
@@ -198,3 +204,16 @@ CREATE TABLE notifications (
     read_at TIMESTAMP WITH TIME ZONE,
     created_at TIMESTAMP WITH TIME ZONE DEFAULT NOW()
 );
+
+CREATE TABLE user_tokens (
+    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+    user_id UUID NOT NULL REFERENCES users(id) ON DELETE CASCADE,
+    name VARCHAR(255) NOT NULL,
+    key_prefix VARCHAR(16) NOT NULL,
+    token_hash TEXT NOT NULL UNIQUE,
+    scope VARCHAR(20) NOT NULL DEFAULT 'write',
+    expires_at TIMESTAMP WITH TIME ZONE,
+    last_used_at TIMESTAMP WITH TIME ZONE,
+    revoked_at TIMESTAMP WITH TIME ZONE,
+    created_at TIMESTAMP WITH TIME ZONE DEFAULT NOW()
+);

app/frontend/src/features/secondary/pages.tsx

@@ -17,6 +17,9 @@ import {
   createAgentToken,
   listAgentTokens,
   revokeAgentToken,
+  createUserToken,
+  listUserTokens,
+  revokeUserToken,
   listDatabases,
   databaseAction,
   updateDatabaseBackupSchedule,
@@ -46,6 +49,8 @@ import {
   updateVulnerability,
   getSecurityMetrics,
   type AgentAuthTokenCreated,
+  type UserTokenCreated,
+  type UserTokenScope,
   type CreateDatabaseInput,
   type DatabaseEntity,
   type FailoverPolicy,
@@ -1267,6 +1272,164 @@ function PlatformSettingsSection() {
   );
 }
 
+const PAT_EXPIRY_OPTIONS: { label: string; days: number }[] = [
+  { label: 'Never', days: 0 },
+  { label: '30 days', days: 30 },
+  { label: '90 days', days: 90 },
+  { label: '1 year', days: 365 },
+];
+
+function UserTokensSection({ isAdmin }: { isAdmin: boolean }) {
+  const queryClient = useQueryClient();
+  const tokensQuery = useQuery({
+    queryKey: ['user-tokens'],
+    queryFn: listUserTokens,
+  });
+  const [tokenName, setTokenName] = useState('');
+  const [tokenScope, setTokenScope] = useState<UserTokenScope>('write');
+  const [expiryDays, setExpiryDays] = useState(0);
+  const [issuedToken, setIssuedToken] = useState<UserTokenCreated | null>(null);
+
+  const createMutation = useMutation({
+    mutationFn: () =>
+      createUserToken({
+        name: tokenName.trim(),
+        scope: tokenScope,
+        expiresInDays: expiryDays,
+      }),
+    onSuccess: (data) => {
+      setIssuedToken(data);
+      setTokenName('');
+      queryClient.invalidateQueries({ queryKey: ['user-tokens'] });
+    },
+  });
+  const revokeMutation = useMutation({
+    mutationFn: (id: string) => revokeUserToken(id),
+    onSuccess: () => queryClient.invalidateQueries({ queryKey: ['user-tokens'] }),
+  });
+
+  const tokens = tokensQuery.data ?? [];
+  const scopeOptions: { value: UserTokenScope; label: string }[] = isAdmin
+    ? [
+        { value: 'read', label: 'Read - GET requests only' },
+        { value: 'write', label: 'Write - read and manage resources' },
+        { value: 'admin', label: 'Admin - everything, incl. platform settings' },
+      ]
+    : [
+        { value: 'read', label: 'Read - GET requests only' },
+        { value: 'write', label: 'Write - read and manage resources' },
+      ];
+
+  return (
+    <section className="panel p-6">
+      <div className="flex items-center gap-3 mb-4">
+        <div className="w-10 h-10 rounded-[var(--radius-md)] bg-[var(--accent-primary-soft)] flex items-center justify-center">
+          <Key size={18} className="text-[var(--accent-primary)]" />
+        </div>
+        <div>
+          <h2 className="text-lg font-semibold text-[var(--text-primary)]">Personal Access Tokens</h2>
+          <p className="text-xs text-[var(--text-tertiary)]">Authenticate the CLI, MCP server, and scripts - same rights as you, bounded by scope</p>
+        </div>
+      </div>
+
+      <div className="space-y-3">
+        <div className="flex flex-wrap gap-2">
+          <input
+            value={tokenName}
+            onChange={(e) => setTokenName(e.target.value)}
+            placeholder="Token name (e.g. laptop cli)"
+            className="h-9 flex-1 min-w-[180px] rounded-[var(--radius-md)] border border-[var(--border-subtle)] bg-[var(--surface-muted)] px-3 text-sm text-[var(--text-primary)] placeholder:text-[var(--text-muted)] focus:border-[var(--accent-primary)] transition-all"
+          />
+          <select
+            value={tokenScope}
+            onChange={(e) => setTokenScope(e.target.value as UserTokenScope)}
+            className="h-9 rounded-[var(--radius-md)] border border-[var(--border-subtle)] bg-[var(--surface-muted)] px-3 text-sm text-[var(--text-primary)] focus:border-[var(--accent-primary)] transition-all"
+          >
+            {scopeOptions.map((option) => (
+              <option key={option.value} value={option.value}>{option.label}</option>
+            ))}
+          </select>
+          <select
+            value={expiryDays}
+            onChange={(e) => setExpiryDays(Number(e.target.value))}
+            className="h-9 rounded-[var(--radius-md)] border border-[var(--border-subtle)] bg-[var(--surface-muted)] px-3 text-sm text-[var(--text-primary)] focus:border-[var(--accent-primary)] transition-all"
+          >
+            {PAT_EXPIRY_OPTIONS.map((option) => (
+              <option key={option.days} value={option.days}>{option.label}</option>
+            ))}
+          </select>
+          <button
+            onClick={() => createMutation.mutate()}
+            disabled={createMutation.isPending || !tokenName.trim()}
+            className="h-9 px-4 rounded-[var(--radius-md)] text-sm font-medium text-[var(--accent-on)] disabled:opacity-50 disabled:cursor-not-allowed transition-all"
+            style={{ background: 'var(--accent-primary)' }}
+          >
+            {createMutation.isPending ? 'Creating…' : 'Create token'}
+          </button>
+        </div>
+
+        {issuedToken ? (
+          <div className="rounded-[var(--radius-md)] border border-[var(--success)]/30 bg-[var(--success-soft)] p-3">
+            <div className="flex items-center justify-between gap-2">
+              <span className="mono text-xs text-[var(--text-primary)] break-all">{issuedToken.token}</span>
+              <button
+                onClick={() => void navigator.clipboard.writeText(issuedToken.token ?? '')}
+                className="flex items-center gap-1 text-xs text-[var(--accent-primary)] hover:underline shrink-0"
+              >
+                <Copy size={11} />
+                Copy
+              </button>
+            </div>
+            <p className="mt-1 text-xs text-[var(--text-tertiary)]">
+              Shown once - store it now. Use with <span className="mono">containr auth login --token</span> or as a Bearer token.
+            </p>
+          </div>
+        ) : null}
+        {createMutation.isError ? (
+          <p className="text-xs text-[var(--error)]">
+            {createMutation.error instanceof Error ? createMutation.error.message : 'Failed to create token'}
+          </p>
+        ) : null}
+
+        {tokensQuery.isLoading ? (
+          <div className="py-4 text-center">
+            <Loader2 size={16} className="animate-spin mx-auto text-[var(--text-tertiary)]" />
+          </div>
+        ) : tokens.length === 0 ? (
+          <p className="text-xs text-[var(--text-tertiary)]">No tokens yet. Create one to connect the CLI or an agent.</p>
+        ) : (
+          <div className="space-y-1.5 border-t border-[var(--border-subtle)] pt-3">
+            {tokens.map((token) => (
+              <div key={token.id} className="flex items-center justify-between gap-2 text-xs">
+                <div className="min-w-0">
+                  <span className="text-[var(--text-primary)]">{token.name}</span>
+                  <span className="ml-2 mono text-[var(--text-muted)]">{token.key_prefix}…</span>
+                  <span className="ml-2 rounded px-1.5 py-0.5 bg-[var(--surface-muted)] text-[var(--text-secondary)]">{token.scope}</span>
+                  <span className="ml-2 text-[var(--text-muted)]">
+                    {token.last_used_at ? `used ${formatRelative(token.last_used_at)}` : 'unused'}
+                    {token.expires_at ? ` · expires ${formatRelative(token.expires_at)}` : ''}
+                  </span>
+                </div>
+                <button
+                  onClick={() => {
+                    if (window.confirm(`Revoke token "${token.name}"? Clients using it lose access immediately.`)) {
+                      revokeMutation.mutate(token.id ?? '');
+                    }
+                  }}
+                  disabled={revokeMutation.isPending}
+                  className="text-[var(--error)] hover:underline disabled:opacity-50 shrink-0"
+                >
+                  Revoke
+                </button>
+              </div>
+            ))}
+          </div>
+        )}
+      </div>
+    </section>
+  );
+}
+
 export function SettingsPage() {
   const navigate = useNavigate();
   const queryClient = useQueryClient();
@@ -1438,6 +1601,9 @@ export function SettingsPage() {
             )}
           </section>
 
+          {/* Personal Access Tokens */}
+          <UserTokensSection isAdmin={profileQuery.data?.isAdmin ?? false} />
+
           {/* Platform Settings - owner only */}
           {profileQuery.data?.isAdmin ? <PlatformSettingsSection /> : null}
 

app/frontend/src/generated/api-types.ts

@@ -231,6 +231,153 @@ export interface paths {
         patch?: never;
         trace?: never;
     };
+    "/user/tokens": {
+        parameters: {
+            query?: never;
+            header?: never;
+            path?: never;
+            cookie?: never;
+        };
+        /**
+         * List personal access tokens
+         * @description Returns the caller's non-revoked personal access token metadata. Token values and hashes are never returned.
+         */
+        get: {
+            parameters: {
+                query?: never;
+                header?: never;
+                path?: never;
+                cookie?: never;
+            };
+            requestBody?: never;
+            responses: {
+                /** @description Token list */
+                200: {
+                    headers: {
+                        [name: string]: unknown;
+                    };
+                    content: {
+                        "application/json": {
+                            tokens?: components["schemas"]["UserToken"][];
+                        };
+                    };
+                };
+                /** @description Unauthorized */
+                401: {
+                    headers: {
+                        [name: string]: unknown;
+                    };
+                    content: {
+                        "application/json": components["schemas"]["Error"];
+                    };
+                };
+            };
+        };
+        put?: never;
+        /**
+         * Create personal access token
+         * @description Issues a `cnp_` bearer token for CLI/MCP/agent authentication. The raw token is returned exactly once and never stored — only its SHA-256 hash persists. Admin-scope tokens require a platform admin account.
+         */
+        post: {
+            parameters: {
+                query?: never;
+                header?: never;
+                path?: never;
+                cookie?: never;
+            };
+            requestBody: {
+                content: {
+                    "application/json": components["schemas"]["CreateUserTokenRequest"];
+                };
+            };
+            responses: {
+                /** @description Token created; raw value returned once */
+                201: {
+                    headers: {
+                        [name: string]: unknown;
+                    };
+                    content: {
+                        "application/json": components["schemas"]["UserTokenCreated"];
+                    };
+                };
+                /** @description Invalid input */
+                400: {
+                    headers: {
+                        [name: string]: unknown;
+                    };
+                    content: {
+                        "application/json": components["schemas"]["Error"];
+                    };
+                };
+                /** @description Admin scope requires platform admin */
+                403: {
+                    headers: {
+                        [name: string]: unknown;
+                    };
+                    content: {
+                        "application/json": components["schemas"]["Error"];
+                    };
+                };
+            };
+        };
+        delete?: never;
+        options?: never;
+        head?: never;
+        patch?: never;
+        trace?: never;
+    };
+    "/user/tokens/{id}": {
+        parameters: {
+            query?: never;
+            header?: never;
+            path?: never;
+            cookie?: never;
+        };
+        get?: never;
+        put?: never;
+        post?: never;
+        /**
+         * Revoke personal access token
+         * @description Marks one of the caller's tokens as revoked. Clients presenting it are rejected from then on.
+         */
+        delete: {
+            parameters: {
+                query?: never;
+                header?: never;
+                path: {
+                    id: string;
+                };
+                cookie?: never;
+            };
+            requestBody?: never;
+            responses: {
+                /** @description Token revoked */
+                200: {
+                    headers: {
+                        [name: string]: unknown;
+                    };
+                    content: {
+                        "application/json": {
+                            message?: string;
+                        };
+                    };
+                };
+                /** @description Token not found */
+                404: {
+                    headers: {
+                        [name: string]: unknown;
+                    };
+                    content: {
+                        "application/json": components["schemas"]["Error"];
+                    };
+                };
+            };
+        };
+        options?: never;
+        head?: never;
+        patch?: never;
+        trace?: never;
+    };
     "/projects": {
         parameters: {
             query?: never;
@@ -637,7 +784,15 @@ export interface paths {
                                 desired?: number;
                                 /** @enum {string} */
                                 status?: "running" | "degraded" | "stopped";
+                                /** @description Reachable service URLs; rewritten to the request Host for remote clients. Null/empty when nothing is published. */
                                 urls?: string[];
+                                /** @description Host ports actually bound for replica 0+; persisted on the service so restarts reuse the same public port. Null/empty when nothing is published. */
+                                ports?: number[];
+                                /**
+                                 * @description Result of the backend's HTTP probe of the service on its project network. Omitted when the service declares no healthcheck path.
+                                 * @enum {string}
+                                 */
+                                health?: "healthy" | "unhealthy" | "unreachable";
                                 containers?: {
                                     id?: string;
                                     name?: string;
@@ -8468,6 +8623,37 @@ export interface components {
             /** @description Human-readable label for the token (e.g. node name) */
             label?: string;
         };
+        UserToken: {
+            /** Format: uuid */
+            id: string;
+            name: string;
+            /** @description First 12 chars of the token, for identification in lists */
+            key_prefix: string;
+            /** @enum {string} */
+            scope: "read" | "write" | "admin";
+            /** Format: date-time */
+            expires_at?: string;
+            /** Format: date-time */
+            last_used_at?: string;
+            /** Format: date-time */
+            created_at: string;
+        };
+        UserTokenCreated: components["schemas"]["UserToken"] & {
+            /** @description Raw cnp_ token value, returned exactly once at creation */
+            token: string;
+        };
+        CreateUserTokenRequest: {
+            /** @description Human-readable label (e.g. "laptop cli", "ci deploy") */
+            name: string;
+            /**
+             * @description read = GET/HEAD only; write = all non-admin routes; admin = everything (requires admin account)
+             * @default write
+             * @enum {string}
+             */
+            scope: "read" | "write" | "admin";
+            /** @description 0 or omitted = never expires */
+            expires_in_days?: number;
+        };
         AgentCapabilities: {
             container_runtimes?: string[];
             supported_architectures?: string[];

app/frontend/src/generated/docs-snapshot.ts

@@ -52,7 +52,7 @@ export const bundledDocs: BundledDoc[] = [
     "path": "guides/OPERATIONS.md",
     "section": "guides",
     "title": "Operations",
-    "markdown": "# Operations\n\nDay-2 features for running services: scheduled jobs, ad-hoc exec, managed\ndatabases, preview environments, security scans, and high availability.\n\n## Cron jobs\n\nService detail → **Cron**. Standard 5-field cron expressions (`robfig/cron`\nsyntax), per-job timezone, enable/disable, manual trigger, and execution\nhistory with captured stdout/stderr.\n\n- The scheduler ticks every minute and runs due jobs via `docker exec`\n  (`sh -c <command>`) inside the service's container.\n- Executions are retained per-job (`retention` count) and visible in the UI.\n- Requires a running container — a stopped service's due job records a\n  failure, it does not queue.\n\n```\nGET/POST /api/v1/cron-jobs            PUT/DELETE /api/v1/cron-jobs/:id\nPOST     /api/v1/cron-jobs/:id/trigger\nGET      /api/v1/cron-jobs/:id/executions\n```\n\n## Exec console\n\nService detail → **Console**. One-off `sh -c` commands against the running\ncontainer: 30s ceiling, 64KB output cap, exit code + captured output shown\nper run. Debugging aid — not an interactive shell.\n\n```\nPOST /api/v1/services/:id/exec  {command}\n```\n\n## Managed databases\n\n**/databases** page — provision PostgreSQL/Redis/MySQL/MariaDB/MongoDB/\nClickHouse/Dragonfly as managed containers on the `containr` network.\n\n- Start/stop/restart actions; connection URL with copy.\n- **Backups**: manual snapshot now, or set a cron expression\n  (`backup_schedule`) — due backups run inside the same scheduler tick as\n  cron jobs. Archives live in the `containr-db-backups` volume and are\n  downloadable (`GET /databases/:id/backups/:bid/download`) and restorable.\n- **Bind to service**: injects `connection_url` into a service's variables\n  as a secret (`DATABASE_URL` by default). Redeploy the service afterward —\n  variables are read at deploy time.\n\n## Preview environments\n\nService detail → **Previews**. Create a preview per branch/PR with a TTL,\npromote it to production, or delete it.\n\nCurrent limitation: previews are bookkeeping records — they do not build,\ndeploy, or route a real container. The generated `*.preview.containr.local`\nURL is a placeholder. Real preview runtime is tracked on the roadmap.\n\n## Security scans\n\n**/security** page — pick a project, run a scan (dependency / configuration\n/ comprehensive, optionally scoped to one service).\n\nThe scanner is heuristic config analysis, not Trivy/Grype: it flags\nunpinned image tags, `http://` source/public URLs, `npm install` without\n`npm ci`, `pip install` without hashes, `curl|sh` build steps, missing\nhealth checks, and similar smells. Findings get severity badges and can be\nresolved or ignored; the page also shows a security score, open/resolved\ncounts, scan history, and compliance status.\n\n```\nPOST /api/v1/security/scans     {project_id, service_id?, scan_type}\nGET  /api/v1/projects/:id/vulnerabilities\nPUT  /api/v1/vulnerabilities/:id          {status: open|resolved|ignored}\nGET  /api/v1/projects/:id/security/metrics\nGET  /api/v1/projects/:id/security/history\n```\n\n## High availability\n\n**/ha** page — manager enable/disable, node/health-check/alert status\ncards, manual failover (with confirmation), per-service failover policies\n(project → service picker, strategy, min healthy nodes, max failures),\nactive alerts with resolve, and health check results.\n\nHA state is in-memory like autoscaling — policies and alerts reset on\nbackend restart.\n\n## Node agents\n\n**/usage** page → *Connect node*: issue a per-node token (`cagt_…`, shown\nonce), copy the prefilled install command, revoke tokens when done. Agents\nregister over `POST /api/agents/register` and heartbeat via\n`POST /api/agents/heartbeat`.\n"
+    "markdown": "# Operations\n\nDay-2 features for running services: scheduled jobs, ad-hoc exec, managed\ndatabases, preview environments, security scans, and high availability.\n\n## Cron jobs\n\nService detail → **Cron**. Standard 5-field cron expressions (`robfig/cron`\nsyntax), per-job timezone, enable/disable, manual trigger, and execution\nhistory with captured stdout/stderr.\n\n- The scheduler ticks every minute and runs due jobs via `docker exec`\n  (`sh -c <command>`) inside the service's container.\n- Executions are retained per-job (`retention` count) and visible in the UI.\n- Requires a running container — a stopped service's due job records a\n  failure, it does not queue.\n\n```\nGET/POST /api/v1/cron-jobs            PUT/DELETE /api/v1/cron-jobs/:id\nPOST     /api/v1/cron-jobs/:id/trigger\nGET      /api/v1/cron-jobs/:id/executions\n```\n\n## Exec console\n\nService detail → **Console**. One-off `sh -c` commands against the running\ncontainer: 30s ceiling, 64KB output cap, exit code + captured output shown\nper run. Debugging aid — not an interactive shell.\n\n```\nPOST /api/v1/services/:id/exec  {command}\n```\n\n## Managed databases\n\n**/databases** page — provision PostgreSQL/Redis/MySQL/MariaDB/MongoDB/\nClickHouse/Dragonfly as managed containers on the `containr` network.\n\n- Start/stop/restart actions; connection URL with copy.\n- **Backups**: manual snapshot now, or set a cron expression\n  (`backup_schedule`) — due backups run inside the same scheduler tick as\n  cron jobs. Archives live in the `containr-db-backups` volume and are\n  downloadable (`GET /databases/:id/backups/:bid/download`) and restorable.\n- **Bind to service**: injects `connection_url` into a service's variables\n  as a secret (`DATABASE_URL` by default). Redeploy the service afterward —\n  variables are read at deploy time.\n\n## Preview environments\n\nService detail → **Previews**. Create a preview per branch/PR with a TTL,\npromote it to production, or delete it.\n\nCurrent limitation: previews are bookkeeping records — they do not build,\ndeploy, or route a real container. The generated `*.preview.containr.local`\nURL is a placeholder. Real preview runtime is tracked on the roadmap.\n\n## Security scans\n\n**/security** page — pick a project, run a scan (dependency / configuration\n/ comprehensive, optionally scoped to one service).\n\nThe scanner is heuristic config analysis, not Trivy/Grype: it flags\nunpinned image tags, `http://` source/public URLs, `npm install` without\n`npm ci`, `pip install` without hashes, `curl|sh` build steps, missing\nhealth checks, and similar smells. Findings get severity badges and can be\nresolved or ignored; the page also shows a security score, open/resolved\ncounts, scan history, and compliance status.\n\n```\nPOST /api/v1/security/scans     {project_id, service_id?, scan_type}\nGET  /api/v1/projects/:id/vulnerabilities\nPUT  /api/v1/vulnerabilities/:id          {status: open|resolved|ignored}\nGET  /api/v1/projects/:id/security/metrics\nGET  /api/v1/projects/:id/security/history\n```\n\n## High availability\n\n**/ha** page — manager enable/disable, node/health-check/alert status\ncards, manual failover (with confirmation), per-service failover policies\n(project → service picker, strategy, min healthy nodes, max failures),\nactive alerts with resolve, and health check results.\n\nHA state is in-memory like autoscaling — policies and alerts reset on\nbackend restart.\n\n## Node agents\n\n**/usage** page → *Connect node*: issue a per-node token (`cagt_…`, shown\nonce), copy the prefilled install command, revoke tokens when done. Agents\nregister over `POST /api/agents/register` and heartbeat via\n`POST /api/agents/heartbeat`.\n\n## Personal access tokens\n\n**Settings → Personal Access Tokens**: issue `cnp_…` bearer tokens for the\nCLI, MCP server, and scripts — the same privileges as your account, bounded\nby scope. The raw token is shown once at creation; only its SHA-256 hash is\nstored.\n\nScopes:\n\n- `read` — `GET`/`HEAD`/`OPTIONS` only; mutations return `403 SCOPE_INSUFFICIENT`.\n- `write` — everything except admin routes.\n- `admin` — full access; requires a platform admin account, and the token is\n  still capped at the owner's privileges.\n\nA token can never outrank its owner: non-admin scopes report\n`is_admin = false` even on admin accounts.\n\n```\nGET    /api/v1/user/tokens            list (metadata only, never hashes)\nPOST   /api/v1/user/tokens            {name, scope?, expires_in_days?} → raw token once\nDELETE /api/v1/user/tokens/:id        revoke (owner-scoped)\n```\n\nUse as `Authorization: Bearer cnp_…`, or `containr auth login cnp_…`.\nRevocation takes effect on the next request; `last_used_at` updates on each\nauthenticated call.\n"
   },
   {
     "path": "guides/README_BACKEND.md",

app/frontend/src/index.css

@@ -1523,6 +1523,7 @@ textarea {
   background: var(--accent-primary); color: var(--accent-on);
   border: none; border-radius: 8px; padding: 9px 17px;
   font: 700 12px 'JetBrains Mono', monospace; cursor: pointer;
+  white-space: nowrap; flex-shrink: 0;
   transition: background 0.15s, transform 0.15s;
   box-shadow: 0 2px 12px rgba(180, 227, 74, 0.22);
 }
@@ -1534,6 +1535,7 @@ textarea {
   background: var(--surface-card); color: var(--text-secondary);
   border: 1px solid var(--border-subtle); border-radius: 8px; padding: 8px 16px;
   font: 600 12px 'JetBrains Mono', monospace; cursor: pointer; transition: 0.15s;
+  white-space: nowrap; flex-shrink: 0;
 }
 .v-btn-ghost:hover { color: var(--text-primary); border-color: var(--border-default); }
 

app/frontend/src/lib/api-client.ts

@@ -993,6 +993,34 @@ export async function revokeAgentToken(id: string): Promise<void> {
   await requestJson(`/agent-tokens/${id}`, { method: 'DELETE' });
 }
 
+export type UserToken = components['schemas']['UserToken'];
+export type UserTokenCreated = components['schemas']['UserTokenCreated'];
+export type UserTokenScope = 'read' | 'write' | 'admin';
+
+export async function createUserToken(input: {
+  name: string;
+  scope?: UserTokenScope;
+  expiresInDays?: number;
+}): Promise<UserTokenCreated> {
+  return requestJson<UserTokenCreated>(`/user/tokens`, {
+    method: 'POST',
+    body: JSON.stringify({
+      name: input.name,
+      scope: input.scope ?? 'write',
+      expires_in_days: input.expiresInDays ?? 0,
+    }),
+  });
+}
+
+export async function listUserTokens(): Promise<UserToken[]> {
+  const payload = await requestJson<{ tokens?: UserToken[] }>(`/user/tokens`);
+  return payload.tokens ?? [];
+}
+
+export async function revokeUserToken(id: string): Promise<void> {
+  await requestJson(`/user/tokens/${id}`, { method: 'DELETE' });
+}
+
 export async function createProject(input: CreateProjectInput): Promise<ProjectEntity> {
   const payload = await requestJson<RawProject | { project?: RawProject }>(`/projects`, {
     method: 'POST',

docs/api/openapi.yaml

@@ -152,6 +152,97 @@ paths:
               schema:
                 $ref: '#/components/schemas/Error'
 
+  /user/tokens:
+    get:
+      tags:
+        - User
+      summary: List personal access tokens
+      description: Returns the caller's non-revoked personal access token metadata. Token values and hashes are never returned.
+      security:
+        - bearerAuth: []
+      responses:
+        '200':
+          description: Token list
+          content:
+            application/json:
+              schema:
+                type: object
+                properties:
+                  tokens:
+                    type: array
+                    items:
+                      $ref: '#/components/schemas/UserToken'
+        '401':
+          description: Unauthorized
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/Error'
+    post:
+      tags:
+        - User
+      summary: Create personal access token
+      description: Issues a `cnp_` bearer token for CLI/MCP/agent authentication. The raw token is returned exactly once and never stored — only its SHA-256 hash persists. Admin-scope tokens require a platform admin account.
+      security:
+        - bearerAuth: []
+      requestBody:
+        required: true
+        content:
+          application/json:
+            schema:
+              $ref: '#/components/schemas/CreateUserTokenRequest'
+      responses:
+        '201':
+          description: Token created; raw value returned once
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/UserTokenCreated'
+        '400':
+          description: Invalid input
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/Error'
+        '403':
+          description: Admin scope requires platform admin
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/Error'
+
+  /user/tokens/{id}:
+    delete:
+      tags:
+        - User
+      summary: Revoke personal access token
+      description: Marks one of the caller's tokens as revoked. Clients presenting it are rejected from then on.
+      security:
+        - bearerAuth: []
+      parameters:
+        - name: id
+          in: path
+          required: true
+          schema:
+            type: string
+            format: uuid
+      responses:
+        '200':
+          description: Token revoked
+          content:
+            application/json:
+              schema:
+                type: object
+                properties:
+                  message:
+                    type: string
+        '404':
+          description: Token not found
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/Error'
+
   # Projects
   /projects:
     get:
@@ -458,6 +549,16 @@ paths:
                         type: array
                         items:
                           type: string
+                        description: Reachable service URLs; rewritten to the request Host for remote clients. Null/empty when nothing is published.
+                      ports:
+                        type: array
+                        items:
+                          type: integer
+                        description: Host ports actually bound for replica 0+; persisted on the service so restarts reuse the same public port. Null/empty when nothing is published.
+                      health:
+                        type: string
+                        enum: [healthy, unhealthy, unreachable]
+                        description: Result of the backend's HTTP probe of the service on its project network. Omitted when the service declares no healthcheck path.
                       containers:
                         type: array
                         items:
@@ -6385,6 +6486,66 @@ components:
           type: string
           description: Human-readable label for the token (e.g. node name)
 
+    UserToken:
+      type: object
+      properties:
+        id:
+          type: string
+          format: uuid
+        name:
+          type: string
+        key_prefix:
+          type: string
+          description: First 12 chars of the token, for identification in lists
+        scope:
+          type: string
+          enum: [read, write, admin]
+        expires_at:
+          type: string
+          format: date-time
+        last_used_at:
+          type: string
+          format: date-time
+        created_at:
+          type: string
+          format: date-time
+      required:
+        - id
+        - name
+        - key_prefix
+        - scope
+        - created_at
+
+    UserTokenCreated:
+      allOf:
+        - $ref: '#/components/schemas/UserToken'
+        - type: object
+          properties:
+            token:
+              type: string
+              description: Raw cnp_ token value, returned exactly once at creation
+          required:
+            - token
+
+    CreateUserTokenRequest:
+      type: object
+      properties:
+        name:
+          type: string
+          description: Human-readable label (e.g. "laptop cli", "ci deploy")
+        scope:
+          type: string
+          enum: [read, write, admin]
+          default: write
+          description: read = GET/HEAD only; write = all non-admin routes; admin = everything (requires admin account)
+        expires_in_days:
+          type: integer
+          minimum: 0
+          maximum: 3650
+          description: 0 or omitted = never expires
+      required:
+        - name
+
     AgentCapabilities:
       type: object
       properties:

docs/guides/OPERATIONS.md

@@ -90,3 +90,30 @@ backend restart.
 once), copy the prefilled install command, revoke tokens when done. Agents
 register over `POST /api/agents/register` and heartbeat via
 `POST /api/agents/heartbeat`.
+
+## Personal access tokens
+
+**Settings → Personal Access Tokens**: issue `cnp_…` bearer tokens for the
+CLI, MCP server, and scripts — the same privileges as your account, bounded
+by scope. The raw token is shown once at creation; only its SHA-256 hash is
+stored.
+
+Scopes:
+
+- `read` — `GET`/`HEAD`/`OPTIONS` only; mutations return `403 SCOPE_INSUFFICIENT`.
+- `write` — everything except admin routes.
+- `admin` — full access; requires a platform admin account, and the token is
+  still capped at the owner's privileges.
+
+A token can never outrank its owner: non-admin scopes report
+`is_admin = false` even on admin accounts.
+
+```
+GET    /api/v1/user/tokens            list (metadata only, never hashes)
+POST   /api/v1/user/tokens            {name, scope?, expires_in_days?} → raw token once
+DELETE /api/v1/user/tokens/:id        revoke (owner-scoped)
+```
+
+Use as `Authorization: Bearer cnp_…`, or `containr auth login cnp_…`.
+Revocation takes effect on the next request; `last_used_at` updates on each
+authenticated call.