Golem

Golem watches every service I run. It reads their logs, traces and metrics from Sensorium, probes their URLs and checks disks and backups.

When something breaks, it opens an incident and sends a short explanation to Telegram and to signed webhooks. When the problem goes away, it closes the incident by itself.

For a known fix, it writes a plan, asks for approval in Telegram, runs the fix and checks that the alert has cleared. Agents add new rules over MCP.