Never wake up to a dead agent again
The OpenClaw gateway is the nerve center of your entire AI operation. When it's down, nothing works. Messages don't route. Crons don't fire. Automations stop. Agents can't reach you. In our experience, the gateway doesn't announce when it goes down — it just silently stops responding while you assume everything is fine.
We built the Gateway Watchdog after losing 13 hours of uptime because nobody noticed the gateway was down. Thirteen hours of missed messages, failed crons, and automations that never ran — all because there was no system watching the watcher. That motivated us to build a proper watchdog that runs independently of the gateway itself.
The watchdog runs every 5 minutes via a macOS LaunchAgent. The health check logic hits the gateway's health endpoint. If it's healthy, the watchdog logs a status line and exits. If it's unhealthy, the watchdog runs openclaw doctor --fix first to resolve configuration issues before attempting a restart. If doctor fixes the issue, it verifies recovery. If the gateway still won't come up, it restarts and verifies again.
Every restart is logged with a timestamp, the failure reason, and the recovery action taken. The log accumulates over time as a diagnostic tool — if you see restart patterns, you can identify what's causing instability. Alert configuration lets you receive a Telegram notification every time the watchdog has to intervene.
Consecutive failure detection adds a cooldown period between restart attempts, preventing the watchdog from hammering a gateway that's genuinely broken. After N consecutive failures, it escalates the alert so you know it needs manual attention rather than an automatic fix.
This is the most boring product on this marketplace. It's also one of the most important. Set it up once, forget about it, and wake up every morning to a gateway that's been quietly restarting itself whenever needed. Simple, critical, battle-tested since March 2026.