MCP status monitoring for AI agents

AI coding agents are good at scaffolding routes, env vars, and deploy configs. They are worse at remembering to log into a dashboard after ship night and add the production health URL. Model Context Protocol (MCP) closes that gap: your editor exposes tools the model can call with your permission — list projects, create a check, fetch public status — without pasting secrets into chat.

stillonline-mcp is StillOnline’s npm package. It wraps our REST API v1 for Pro and Ultimate accounts. It is not a separate product; it is how Cursor, Claude Desktop, and other MCP clients automate the same objects you manage at stillonline.tech/en/app.

What you get

Published as stillonline-mcp. Typical tools:

• list_projects — see slugs and IDs before creating checks.
• create_check — project_id, url, optional name and interval.
• list_checks — verify what the agent created.
• get_public_status — read the customer-facing page JSON by slug (no API key on the public endpoint).
• create_incident / list_incidents — comms when you are already in incident mode.

Full reference: /en/docs/mcp. OpenAPI base: https://api.stillonline.tech/v1.

Setup in five minutes

1. Upgrade or start on Pro ($9/mo) if you need API/MCP — Free has no API key. See Pricing.
2. Create sk_live_… at API settings (shown once).
3. Add MCP config (never put the key in the prompt — only in env):

{
  "mcpServers": {
    "stillonline": {
      "command": "npx",
      "args": ["-y", "stillonline-mcp"],
      "env": {
        "STILLONLINE_API_KEY": "sk_live_..."
      }
    }
  }
}

1. Restart MCP in Cursor (Settings → MCP) or Claude Desktop (Developer config).
2. Ask the agent: “Create an HTTP check for https://api.example.com/health on my StillOnline project X.”

Optional env: STILLONLINE_API_BASE if you proxy the API (default is production).

Good prompts (and bad ones)

Good: “List my StillOnline projects, then add a 5-minute HTTP check expecting 200 to /api/health on project my-saas.”

Bad: “Here is my API key sk_live_… — store it in memory.” Keys belong in MCP env only.

MCP shines after you add a health route in code — pair with Vercel monitoring or Railway. For OpenClaw gateways, you still need a public health URL first (OpenClaw post).

Honest limits

MCP does not replace external HTTP probes — it configures them. StillOnline is not Datadog: no log tailing, no trace UI. Compare Better Stack if you need observability suites.

Security recap from our docs: StillOnline and npm never ask for your key during npx install. Rotate keys if one leaks into a shared transcript.

FAQ

Does MCP work on Free?

No. API keys and stillonline-mcp require Pro or Ultimate. Free users use the web UI only.

Cursor vs Claude Desktop — same package?

Yes. Any MCP client that runs npx -y stillonline-mcp with STILLONLINE_API_KEY in env.

Can the agent read private status pages?

get_public_status targets public slugs. Private pages stay behind your app auth on Pro+.

Will the agent create 100 checks by mistake?

Pro caps at 10 URLs. Ultimate at 25. Prompt with a single target URL to stay safe.

Is this related to OpenClaw MCP?

Different servers. OpenClaw has its own gateway tools; stillonline-mcp only talks to StillOnline’s API for uptime and incidents.