Agent Production Monitoring Kit

Four Python scripts that monitor a self-hosted AI agent system and post Slack alerts when something needs attention. Silent when everything is running normally.

The four scripts: a log scanner (hourly) that parses OpenClaw logs for error patterns and sends a Slack message listing what went wrong. A cron watchdog (every 15 minutes) that checks heartbeat files written by your other cron jobs: if a heartbeat has not been updated within its expected window, you get an alert. An API usage monitor (daily) that compares yesterday spend against a 7-day rolling average and fires at 2.5x threshold. A secret scanner (weekly) using TruffleHog, checking your codebase for exposed credentials.

On the first run of the secret scanner against our own system: caught a hardcoded API key in a shell script.

What is included: all four Python scripts ready to deploy, cron setup instructions for Ubuntu, Slack webhook integration guide, and the heartbeat file pattern that makes the watchdog work.

Best for operators self-hosting AI systems on a VPS who want to know about problems before their users do.