Silentwatch Mcp

Catches the class of scheduled job failures that traditional monitoring misses: exit code 0 but empty output, overdue runs nobody noticed, jobs "succeeding" at 10% rate while dashboards stay green. Exposes six MCP tools (list_jobs, get_job_status, find_overdue_jobs, find_silent_failures, tail_job_logs) that query crontab, systemd timers, and OpenClaw JSONL logs directly. Ships with configurable backends so you can point it at whatever scheduler you actually run. Designed for the operator running scheduled jobs on a small VPS where full observability stacks are overkill but you still need to know when your backup cron wrote a zero-byte file three days ago. The demo mode shows detection working against synthetic failure patterns in thirty seconds.