heartline

A self-hosted status page for the services you run yourself.

Uptime graphs, incidents and maintenance for your visitors. Monitors, alerts, software updates and an MCP server for you. One container, one YAML file.

What you get

For your visitors

• 📈 Uptime at a glance: a graph per service over 7, 30, 90 days or any date range, with outages in red and degraded time in amber. A timeline view shows one bar per day instead.
• 📣 Incidents and maintenance: open incidents on top, a history with every update, scheduled windows, an RSS feed and CSV reports.
• 🎨 Five colour themes, dark and light, on desktop and phone.

For you

• 🩺 Pull-based checks: HTTP, TCP, DNS, TLS certificates, ICMP and exec plugins for anything else (backup age, queue length, a scripted login). Your services are never modified.
• 🌐 Third-party services: follow Atlassian, GitHub, Google Workspace, Slack, Azure or your Microsoft 365 tenant from their own status pages, shown apart from your services and linked to them with dependsOn.
• 🛠️ One sidebar for everything once you sign in: incidents, monitors, maintenance, notifications, API tokens and users, next to every service.
• 🔔 Alerts to Microsoft Teams, e-mail and webhooks, with recovery notices and reminders.
• 🛡️ Software inventory: the running version of every service, the latest release, end of life and known vulnerabilities (OSV.dev, NVD, CISA KEV), scored from none to critical.
• 🔑 Sign-in with local accounts or Gitea / any OIDC provider, roles from groups, scoped API tokens.
• 🤖 MCP server at /mcp: ask your assistant what is down, declare incidents, schedule maintenance.
• 🗄️ SQLite or Postgres; with Postgres, run several replicas and one leader runs the checks.

Every service: uptime, response times, incidents	Incidents with public updates

Quick start

docker run -d --name heartline -p 3000:3000 \
  -e BASE_URL=http://localhost:3000 \
  -e AUTH_SECRET="$(openssl rand -hex 24)" \
  -e ADMIN_PASSWORD=choose-a-long-password \
  -v heartline-data:/data \
  ghcr.io/bircni/heartline:latest

Open http://localhost:3000. Without a config file heartline monitors itself; sign in at /login as admin@example.com (or ADMIN_EMAIL) with your ADMIN_PASSWORD, see deploy/docker/status.yaml.

Then describe your own services in a status.yaml and mount it with -v "$PWD/status.yaml:/config/status.yaml:ro". It reloads on every change:

site: { title: Homelab status, timezone: Europe/Berlin }

groups:
  - name: Code
    monitors:
      - { id: gitea, name: Gitea, type: http, url: https://git.example.com/api/healthz }
      - { id: postgres, name: Database, type: tcp, host: db.example.com, port: 5432 }

notifiers:
  - { id: teams, type: teams, webhookUrl: '${TEAMS_WEBHOOK_URL}' }

auth:
  users:
    - { email: you@example.com, name: You, password: '${ADMIN_PASSWORD}', role: admin }

status.example.yaml shows every option.

Kubernetes

kubectl apply -k deploy/kustomize/overlays/sqlite     # one replica, SQLite on a PVC
kubectl apply -k deploy/kustomize/overlays/postgres   # two replicas, DATABASE_URL from the secret

Create the heartline Secret first (example). Run more than one replica only with Postgres. Ping checks as non-root need the pod sysctl net.ipv4.ping_group_range: "0 2147483647". The container runs as UID 1000 with a read-only root filesystem.

Connect your assistant

claude mcp add --transport http status https://status.example.com/mcp

The client signs in through heartline (Gitea or a local account) and gets at most your role.

Documentation

Topic	What's inside
Configuration	status.yaml: site, monitors and their check types, notifiers, env vars
Software versions and vulnerabilities	Version probes, release sources, data sources, urgency scoring, privacy
Subscribers	E-mail sign-up, double opt-in, what is sent, feeds
Sign-in, roles and MCP	Local accounts, Gitea/OIDC with role mapping, MCP tools by scope
HTTP API	Public, software and admin routes; OpenAPI at /api/v1/openapi.json
Plugins	Write a check in any language: exit code, JSON line, arguments
Development and releasing	Local setup, tests, project layout, cutting a release with pnpm release

Development

Requires Node 26 and pnpm 12.

pnpm install
pnpm dev     # API on :3000 + UI on :5173, admin: admin@status.local / change-me-please
pnpm check   # lint, format, typecheck, test and build

More in docs/development.md.

License

MIT, see LICENSE.