I’m building Relia, a monitoring agent for founders and small teams who don’t have an SRE. It ships an MCP server, so you can ask about production from Cursor’s chat instead of opening dashboards.
What you can ask from Cursor:
- “What’s broken in production right now?” (relia_status, relia_list_incidents)
-
- “How do I fix this incident?” (relia_get_fix_plan gives concrete steps; Cursor can then make the change in your code, with your approval)
-
- “Did my last deploy break anything?” (relia_changes)
-
- Slow and failing queries and N+1 (relia_database), one request end to end (relia_trace), your app’s error logs (relia_logs), uptime and checks (relia_uptime)
- It reads your project’s data. The only things it can change are things like marking a fix as deployed, acknowledging an incident or recording a deploy, and Cursor asks you before each one.
- Setup:
- npx -y relia-sdk@latest mcp install
- That adds Relia to Cursor (and any other AI tools it finds). Restart Cursor; the first time, a browser opens so you sign in and choose which projects it can see. No token to copy. You can also add it by hand with the MCP address shown on your Relia profile.
Pricing: the Free plan gets MCP status and headlines; Growth ($19/mo) and Pro ($49/mo) get all the tools. New accounts get 14 days of Pro, no card.
If you want to see the product first, there’s a demo with no sign-up: Relia - Observability Agent
I’d like to know which questions you’d actually ask from the editor, and what’s missing.