Openuptime MCP Server

OpenUptime

Open-source uptime monitoring and status pages. Runs on Cloudflare Workers with Postgres, or in one Docker container.

Hosted version · Docs · Reference · MIT license

• HTTP, TCP, DNS, heartbeat, PostgreSQL, MySQL and Redis checks, with keyword matching and slow-response alerts
• Incidents, maintenance windows and public status pages with email / RSS subscriptions
• Alerts to Slack, Lark / Feishu, Telegram, PagerDuty, webhooks and email
• Workspaces, projects, roles, API keys, an MCP server for agents, JSON import / export

Deploy to Cloudflare

The site is two Workers on one domain:

Keeping the API under /api/* of the same domain means session cookies stay same-origin. You need no separate API hostname.

Before you start

• A Cloudflare account and a domain whose DNS is on Cloudflare (the zone).
• A Postgres database reachable from the internet (Neon, Supabase, RDS, your own). Workers connect through Hyperdrive.
• Node.js 22+ and pnpm (corepack enable provides the pinned version).
• Workers Paid plan for the Queue (default). On the Free plan, run in inline mode instead: see Reference.

1. Get the code and sign in

git clone https://github.com/becoolme/openuptime.git && cd openuptime
pnpm install
cd apps/api && npx wrangler login

All npx wrangler … commands below run in apps/api unless noted.

2. Create the Queue and Hyperdrive

npx wrangler queues create openuptime-checks
npx wrangler hyperdrive create openuptime-prod --connection-string="postgres://user:pass@host:5432/db"

The second command prints an id. In apps/api/wrangler.jsonc uncomment the hyperdrive line and paste it:

"hyperdrive": [{ "binding": "HYPERDRIVE", "id": "<id printed above>" }],

The commented placement line is optional (it runs the Worker next to your database host); leave it commented unless you want that.

3. Create the schema

From the repository root:

DATABASE_URL="postgres://user:pass@host:5432/db" pnpm db:migrate

It only applies what is new, so run it again after every upgrade. It never issues CREATE SCHEMA, so a role without database-level CREATE works too.

4. Point the config at your domain

Edit two files. In both, replace status.example.com with your own domain.

apps/api/wrangler.jsonc, where zone_name is the zone (the apex domain) even when you serve from a subdomain. Uncomment routes and set:

"routes": [{ "pattern": "status.example.com/api/*", "zone_name": "example.com" }],
"vars": {
  "BETTER_AUTH_URL": "https://status.example.com",   // public origin: cookies, OAuth, email links
  "WEB_ORIGIN": "https://status.example.com",        // same value
  "EMAIL_FROM": "alerts@example.com"                 // sender address; optional, see below
}

Email is optional: without it invitations and alerts are written to the Worker log. To send it, onboard your domain under Email Sending in the Cloudflare dashboard first (see Reference).

apps/web/wrangler.jsonc:

"routes": [{ "pattern": "status.example.com", "custom_domain": true }],
"vars": { "PRODUCT_URL": "https://status.example.com" }   // target of the "Powered by" link on status pages

The custom domain creates its DNS record and certificate for you; the hostname must not already have a DNS record pointing elsewhere.

No domain on Cloudflare? Leave routes out of both files and set BETTER_AUTH_URL and WEB_ORIGIN to the web Worker's https://openuptime-web.<your-subdomain>.workers.dev address. The web Worker then forwards /api/* to the API Worker through the service binding, so cookies stay same-origin.

5. Set the secrets

npx wrangler secret put BETTER_AUTH_SECRET     # openssl rand -hex 32
npx wrangler secret put CRON_SECRET            # any long random string; guards /internal/*

If wrangler asks to create the Worker first, answer yes. Optional: GITHUB_CLIENT_ID and GITHUB_CLIENT_SECRET add "Sign in with GitHub".

6. Deploy

pnpm exec wrangler deploy                     # in apps/api
cd ../web && pnpm build && pnpm exec wrangler deploy

Deploy the API first: the web Worker binds to it by name.

7. Check it

curl https://status.example.com/api/health    # {"ok":true}

Open the domain and sign up. Once your own account exists you can set "DISABLE_SIGNUP": "true" in the API vars and deploy again to close registration (invited people can still join).

Upgrade later: pull, pnpm install, pnpm db:migrate, then deploy the API and the web Worker, in that order.

Self-host with Docker

One container runs the API, the web app and the check scheduler; Postgres comes with the compose file. Either run the published image or build it from source; the settings are the same.

Option A: run the published image

Tagged releases are published to ghcr.io/becoolme/openuptime for linux/amd64 and linux/arm64. You need no clone and no build:

mkdir openuptime && cd openuptime
curl -fsSLO https://raw.githubusercontent.com/becoolme/openuptime/main/docker-compose.yml
curl -fsSL https://raw.githubusercontent.com/becoolme/openuptime/main/.env.docker.example -o .env

Edit .env: set POSTGRES_PASSWORD and BETTER_AUTH_SECRET (openssl rand -hex 32), and uncomment OPENUPTIME_IMAGE. Pin a version (ghcr.io/becoolme/openuptime:0.1.0) or follow the newest release (:latest). Then:

docker compose pull && docker compose up -d     # http://localhost:8080

Upgrade: change the tag if you pinned one, then run the same two commands again. Migrations run on start.

Option B: build from source

git clone https://github.com/becoolme/openuptime.git && cd openuptime
cp .env.docker.example .env      # set POSTGRES_PASSWORD and BETTER_AUTH_SECRET (openssl rand -hex 32)
docker compose up -d --build     # http://localhost:8080

Upgrade: git pull, then docker compose up -d --build.

Either way

The schema is migrated on start and the data lives in the pgdata volume. Set PUBLIC_URL when the app is not served from http://localhost:8080. Email is optional; see Reference.

Once your own account exists, set DISABLE_SIGNUP=true in .env and run docker compose up -d to close registration (invited people can still join).

Develop

pnpm install
cp .env.example apps/api/.dev.vars   # edit DATABASE_URL and secrets
DATABASE_URL=postgres://user:pass@localhost:5432/openuptime_dev pnpm db:migrate   # same URL as in .dev.vars
pnpm dev:api    # http://localhost:8787  (checks run inline every 15s)
pnpm dev:web    # http://localhost:4321  (proxies /api to the API)
pnpm typecheck

MCP server

Agents (Claude, Cursor, Codex and any MCP client) can manage monitors, incidents and status pages over Streamable HTTP at /api/mcp, authorizing with OAuth or an API key:

claude mcp add --transport http openuptime https://openuptime.app/api/mcp   # or https://<your-host>/api/mcp

Client setup, scopes and the tool list are in the MCP docs. An agent skill that teaches the common workflows ships in skills/openuptime: npx skills add becoolme/openuptime. The hosted server is listed in the MCP Registry as app.openuptime/mcp; server.json is its entry.

Tests, how checks run, the MCP server, alert channels and the other details are in docs/reference.md.

Open source vs hosted

The monitoring itself is the same code in both. The hosted service at openuptime.app adds plans, billing and a few features that live only in that build; it plugs into this repo through one interface (Edition, see docs/editions.md), so nothing in the core is locked or throttled.

Pick self-hosting if you want to own the data and the infrastructure, or have no limits to worry about. Pick hosted if you would rather not run anything. Both receive the same core updates.

Contributing & security

See CONTRIBUTING.md. Report vulnerabilities privately as described in SECURITY.md, not in public issues.

License

MIT