Overview
Connected agents report health checks to CelerFlow at configurable intervals. Health data feeds into the agent score, dashboard monitoring, and staleness detection.
Health report payload
Each health check includes:
| Field | Description |
|---|
status | healthy, degraded, or unhealthy. |
latency_ms | Network latency to the CelerFlow API. |
checks | Detailed sub-checks (JSON) with per-component status, latency, disk space, token expiry, etc. |
openclaw_version | The version of OpenClaw running the agent. |
Configurable intervals
| Interval | When to use |
|---|
| 5 min | High-value production agents you need to watch closely. |
| 15 min | Active development agents. |
| 30 min (default) | Most agents. |
| 60 min | Low-activity or batch agents. |
Staleness detection
A background pg_cron job marks agents as stale when no health check arrives within 30 days.
Stale agents are not automatically removed. They remain visible in the dashboard until manually disconnected or removed.
Health status indicators
| Indicator | Meaning |
|---|
| 🟢 Healthy | Reporting normally, last check healthy. |
| 🟡 Degraded | Reporting, but some sub-checks failing. |
| 🔴 Unhealthy | Reporting, but critical sub-checks failing. |
| ⚫ Stale | No health report in 30+ days. |
CLI diagnostics
Runs a local diagnostic that checks:
- Authentication status
- Network connectivity to the CelerFlow API
- Bootstrap token validity
- Agent registration status
