Cursor MCP for uptime monitoring after you ship
You fixed the health route and merged. The boring part is opening another tab to see if the external probe is green. stillonline-mcp in Cursor lets the agent call StillOnline tools: list checks, read last_status, pull public status JSON.
This is a workflow guide for Cursor after deploy. Full multi-IDE setup lives in MCP status monitoring for AI agents. REST details: service uptime API.
Quick answer
On Pro, run npx -y stillonline-mcp setup, add sk_live_… to Cursor MCP env (never in chat), then prompt: “List StillOnline checks for project X—is anything not OPERATIONAL?” MCP reads the same API as curl gates in GitHub Actions. It does not replace Telegram owner alerts when you are away from the IDE.
When Cursor + MCP helps
| Moment | Agent can |
|---|---|
| Right after deploy | checks.list → confirm green after 2–3 intervals |
| New microservice URL | Create HTTP check from chat (Pro) |
| Incident triage | status.get → summarize for paste into support |
| Docs drift | Compare README status link vs actual slug |
MCP does not run probes faster than your StillOnline plan interval (5 min on Free).
Minimal Cursor setup
- Pro or Ultimate — API key required (pricing).
- Terminal:
npx -y stillonline-mcp setup(pick Cursor). - Paste
sk_live_…from API settings into MCPenvonly. - Reload Cursor → confirm stillonline server is connected.
Troubleshooting keys and PowerShell: full MCP guide.
Prompts that match uptime work
Copy after you ship:
- “Call StillOnline
checks.liston my production project. Flag any check wherelast_statusis not OPERATIONAL.” - “
status.getfor slugmy-saas—quote overall status and anyactive_incident.” - “We added
https://api.example.com/v2/health—create a GET check expecting 200.”
Do not paste secrets into the prompt. Rotate the key if it leaked into chat history.
What MCP does not do
- Page Speed / PSI — separate tab; Page Speed monitoring.
- Solve antibot on marketing URLs — fix health path (probes and redirects).
- Post maintenance incidents — UI or CI/CD incident API; closing incidents is UI in v1.
Keep phone alerts on: MCP is for when you are already in the editor.
FAQ
Does StillOnline MCP work on StillOnline Free?
No API key on Free, so MCP cannot authenticate. Use the dashboard and Telegram alerts, or upgrade to Pro.
Is Cursor MCP different from the REST API?
Same backend (api.stillonline.tech/v1). MCP is a tool wrapper for agents; CI uses curl (REST guide).
Can the agent fix a red check automatically?
It can suggest URL or status-code changes; you still deploy fixes. MCP reads state—it does not SSH into your servers.
How is this article different from the general MCP post?
This page is Cursor + post-deploy uptime workflows. The MCP agents guide covers Windsurf, Claude, Antigravity, and install depth.
Will MCP notify me on downtime at night?
No push from MCP. Configure Telegram via StillOnline bot for owner alerts.
Can I use MCP and GitHub Actions together?
Yes. Actions gate merges; Cursor helps you inspect state while coding. Same Pro key, different secret stores.
Further reading
MCP status monitoring for AI agents
Step-by-step MCP setup in popular AI agent IDEs to read service uptime, list checks, and fetch public status — wizard, per-client config, Pro API key.
REST API for service uptime and status
GET public status JSON without a key, or list checks with last_status via Bearer auth — curl examples, rate limits, OPERATIONAL vs paused checks, Pro API key.
GitHub Actions uptime check with StillOnline API
Gate deploys with curl on public status JSON or private checks API. GitHub secrets for sk_live, rate limits, and when probes differ from customer view.
Health URL for SaaS — 5-min quickstart
Add a public GET /health or /api/health endpoint, verify it returns 200, and register an external StillOnline check with a status page — Free plan limits included.