Agent not answering

[sage patio]

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

[tall umbra]

First thing: what does openclaw --version show?

Restarting the gateway rules out one class of issue, but if the agent still won’t answer I’d check whether messages are being dropped vs the model run is failing.

Please run these and paste the relevant output/log tail:

openclaw --version
openclaw status
openclaw channels status --probe
openclaw gateway status
openclaw logs --follow

While watching openclaw logs --follow, send the agent a fresh test message. The log line around that message is usually the key: mention/pairing/allowlist drops look very different from provider/model errors.

Common things to look for:

• pairing request / pairing required → sender or channel needs approval
• drop guild message (mention required → group mention gating/routing issue
• blocked / allowlist → channel/user policy filtered it
• provider errors like 401/429/tool schema/etc. → model/auth/runtime issue, not Discord/gateway

Docs for the triage flow: https://docs.openclaw.ai/help/faq and https://docs.openclaw.ai/gateway/troubleshooting

[balmy cypress]

so what would happen if i do openclaw config set channels.telegram.enabled false
openclaw gateway restart
openclaw status

can i text my agent on discord or gateway dashboard after?

[tall umbra]

That part looks encouraging — tasks are clean and sessions exist — but it’s the lower half of the status output, not the gateway/connectivity bit.

Can you paste the lines for Gateway, Gateway service, Channels, and Probes from openclaw status --deep?

Also try a direct local agent run to separate “agent runtime broken” from “Discord delivery broken”:

openclaw agent run --prompt "reply with exactly: ping-ok"

Interpretation:

• If that replies ping-ok: the agent/model works; then we debug Discord channel routing/pairing/mention policy.
• If that hangs/errors: the issue is the agent/model/session runtime; paste that error plus the latest openclaw logs --follow lines.

One more useful check:

openclaw channels status --probe

Paste the Discord line from that if Discord is what you want to use.

[balmy cypress]

what is going on and why did this happen, ive been using openclaw for months and this has never happened

[tall umbra]

This kind of thing can happen after a VPN/Tailscale change, DNS change, firewall/security update, network switch, or stale bind config. Months of working fine makes sense: the config can be okay until the network address it depended on disappears.

Good call rebooting — now we want to re-check from the top, because the old errors may be stale.

Run these in order and paste the first thing that looks bad:

openclaw gateway status
openclaw status --deep
openclaw channels status --probe

Then test whether the agent itself can run outside Discord:

openclaw agent run --prompt "reply with exactly: ping-ok"

How to read it:

The key next split is: can openclaw agent run answer locally? That tells us whether we’re debugging the agent or just Discord delivery.

[balmy cypress]

agent answered on gateway dashboard, i want to explain what just happened and have my agent look into what commands i ran and what exactly happened, give me a prompt for this, include all commands i might have ran and what errors happened

ok telegram is fixed, btw did i expose anything important in this chat? if yes please delete it

please delete any mildly private/local details though: Mac username/hostname, local paths, service PATH, Telegram account labels

[tall umbra]

Done — I deleted the remaining messages I could see that exposed local/mildly private details, including local paths, service-env/path output, specific Telegram account labels, and the Mac hostname/username-style bits.

I re-read the thread afterward; what remains looks like generic troubleshooting commands/context, not secrets or local identifiers.

[balmy cypress]

thanks