My homelab used to wake me up at 2am. Now it handles its own fires.

I’ve been building a system of Claude AI routines — scheduled Python scripts that gather context from my infrastructure, send it to Claude, and take pre-authorized actions like restarting crashed containers, flagging disk issues, and auditing config drift. The whole thing runs for under $0.15/month.

In the new post I walk through four production routines with full code: a daily health digest, a container crash responder, a disk space guardian, and a weekly config drift checker.

Read the full guide →