OpenClaw `channels status --probe`: 5-Minute Diagnosis for Bots That Are Online but Not Replying (2026)

Have you seen this pattern?

• openclaw gateway status looks healthy
• Your Telegram bot appears online
• But messages still get no reply

The common mistake is checking only the service layer, not channel probes.

This guide gives you the practical workflow: validate gateway first, then use channels status --probe to isolate the real channel issue.

One-line takeaway

• openclaw gateway status: answers “Is Gateway alive?”
• openclaw channels status --probe: answers “Is this specific channel actually usable with current credentials/policy?”

For “online but no reply,” the second command often finds the root cause faster.

1) 60-second baseline checks (copy/paste)

openclaw gateway status --deep
openclaw status --deep --timeout 10000

If gateway is unhealthy, fix that first. If gateway is healthy but replies still fail, continue.

2) Core command: probe channel health directly

openclaw channels status --probe

Need shareable output?

openclaw channels status --probe --json

Optional quick parse (if jq is installed):

openclaw channels status --probe --json | jq '.'

Pending confirmation: exact field names may differ across versions, but --probe behavior is consistent for diagnostics. Run openclaw channels status --help on your host for version-specific output.

3) Interpret results with symptom → action mapping

Scenario A: Gateway healthy, one channel probe fails

Meaning: service layer is fine; problem is channel-side (credentials, permissions, routing, policy).

Collect runtime evidence:

openclaw logs --follow

Then verify channel specifics (Telegram example):

• token loaded by runtime user
• webhook/polling conflict (classic 409 Conflict)
• group gating blocks (groupPolicy/groupAllowFrom)

Scenario B: All channels fail probe

Meaning: likely gateway/network/proxy/runtime environment issue.

openclaw gateway status --deep
openclaw status --all
openclaw logs --follow

Prioritize checks:

1. Service is running on the expected host
2. Reverse proxy/tunnel is reachable
3. Runtime env vars match what your daemon sees

Scenario C: Channel probe passes, users still get silence

Meaning: messages may enter channel path but fail at model/API layer.

openclaw logs --follow

Look for:

• invalid model key
• quota exhaustion
• upstream timeout
• no fallback model configured

4) Fast recovery sequence

# 1) Restart gateway
openclaw gateway restart

# 2) Verify service layer
openclaw gateway status --deep

# 3) Verify channel layer
openclaw channels status --probe

# 4) Observe message path
openclaw logs --follow

If handing off to teammates/community, include this evidence pack:

openclaw status --all
openclaw channels status --probe --json

5) Verifiable done criteria

• openclaw gateway status --deep is healthy
• openclaw channels status --probe passes for target channel
• openclaw logs --follow shows complete inbound + response chain for a test message
• For group chats, policy (groupPolicy / groupAllowFrom) matches your intended access model

When all four are true, “online but no reply” incidents drop significantly.

Final note

channels status --probe does not replace status; it removes ambiguity:

• gateway-layer issue?
• channel-layer issue?
• model-layer issue?

Once the layer is identified, resolution usually takes minutes instead of hours.

Related guides

• OpenClaw status vs gateway status: 5-Minute Debug Flow for “Online but No Reply” (2026)
• OpenClaw Telegram Bot Online but Not Replying: Webhook, 409 Conflict, and Permission Fix Checklist (2026)
• OpenClaw Log Troubleshooting Guide: From Error to Root Cause