A repair command is only half the work. The other half is proving that the gateway, model layer, and channel layer all recovered cleanly afterwards.
Quick answer
Run openclaw doctor --fix --yes, then validate the stack with gateway status, model checks, and channel probes.
Command line steps
1. Apply the automatic repair path
The alias --fix maps to the repair flow and is good for quick maintenance runs.
BASH
$openclaw doctor --fix --yes
2. Validate the gateway and model layer
These two commands catch most post-repair regressions immediately.
BASH
$openclaw gateway status --require-rpc
openclaw models status --check
3. Validate the channel layer too
If you use Telegram, WhatsApp, or other channels, probe them explicitly before you close the incident.
BASH
$openclaw channels status --probe
openclaw logs --limit 200
What to check if it still fails
- If doctor reports aggressive repairs were skipped, do not assume the system is fully fixed.
- If the same incident repeats after repair, capture the live logs and compare them to the previous run instead of auto-repairing forever.
- If a channel still fails but the gateway and models are healthy, isolate the channel config instead of re-running full-stack repairs.