dvcrn Gateway Keeper
OS-level watchdog for OpenClaw gateway. Runs outside the gateway process so it survives crashes.
How It Works
- Health check — Runs
openclaw gateway statusevery 15 minutes - Auto-restart — If gateway is down, runs
openclaw gateway start - Recovery signal — Writes
logs/gateway-recovery.jsonwith crash timestamp - Session recovery — HEARTBEAT.md template detects recovery file, prompts agent to check incomplete work
Quick Setup
Install
Run the appropriate install script for your OS:
Windows (PowerShell as Admin):
powershell -ExecutionPolicy Bypass -File "<skill-dir>/scripts/install.ps1"
Linux/macOS:
bash "<skill-dir>/scripts/install.sh"
Uninstall
Windows:
powershell -ExecutionPolicy Bypass -File "<skill-dir>/scripts/uninstall.ps1"
Linux/macOS:
bash "<skill-dir>/scripts/uninstall.sh"
Recovery Protocol
After gateway restart, the agent should check logs/gateway-recovery.json:
{
"crashed_at": "2026-02-26T00:00:00Z",
"restarted_at": "2026-02-26T00:15:00Z",
"restarted_by": "gateway-keeper"
}
Add to HEARTBEAT.md (done automatically by install script):
## Gateway Crash Recovery
If `logs/gateway-recovery.json` exists:
1. Read crash timestamp
2. List all active sessions/sub-agents
3. Check each for incomplete work
4. Resume or retry as needed
5. Delete the recovery file when done
Files
| File | Purpose |
|---|---|
scripts/check-gateway.ps1 |
Windows health check + restart |
scripts/check-gateway.sh |
Linux/macOS health check + restart |
scripts/install.ps1 |
Windows Task Scheduler setup |
scripts/install.sh |
Linux/macOS cron/systemd setup |
scripts/uninstall.ps1 |
Windows cleanup |
scripts/uninstall.sh |
Linux/macOS cleanup |
Customization
Edit check interval by modifying the scheduled task/cron entry. Default: 15 minutes.
To change recovery behavior, edit the HEARTBEAT.md recovery section.