I am getting terminated as a response constantly | Friends of the Crustacean 🦞🤝 | Page 1

trim delta May 12, 2026, 7:34 PM

#

I am getting terminated message on my openclaw constantly

vernal zephyrBOT May 12, 2026, 7:34 PM

#

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

lethal ferryBOT May 12, 2026, 7:34 PM

#

That usually means the agent run is being killed underneath OpenClaw — either the model connection is closing with a bare terminated, or a CLI-backed model is timing out / producing no output and OpenClaw kills it.

First, can you paste these? They redact tokens and are the fastest way to spot it:

openclaw --version
openclaw status --all

Then reproduce one failed message and grab the nearby log lines:

openclaw logs --follow

Things I’d look for in the logs:

All models failed ... terminated

#

CLI produced no output ... and was terminated
CLI exceeded timeout ... and was terminated
gateway restart/crash messages
provider HTTP errors / rate limits

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

Once you paste the version + status/log snippet, we can tell whether this is model/provider, gateway crash, timeout, or install/version mismatch. 🦞

trim delta May 12, 2026, 7:35 PM

#

marcelo@Marcelos-iMac ~ % openclaw --version
OpenClaw 2026.4.22 (00bd2cf)

lethal ferryBOT May 12, 2026, 7:35 PM

#

Thanks — 2026.4.22 is recent enough that I wouldn’t jump straight to “old version” as the cause.

Next thing we need is the failure context. Please run:

openclaw status --all

And then, right after you get the terminated response again:

openclaw logs --follow

Copy the last ~40 lines around the failure. The important bit is whether it says a provider call returned plain terminated, or something like CLI produced no output... and was terminated / CLI exceeded timeout... — those point to different fixes.

trim delta May 12, 2026, 7:37 PM

#

19:35:37+00:00 warn skills {"subsystem":"skills"} {"source":"openclaw-managed","rootDir":"/Users/marcelo/.openclaw/skills","rootRealPath":"/Users/marcelo/.openclaw/skills","path":"/Users/marcelo/.openclaw/skills/ui-ux-pro-max","realPath":"/Users/marcelo/.agents/skills/ui-ux-pro-max","reason":"symlink-escape"} Skipping escaped skill path outside its configured root.
19:35:37+00:00 warn skills {"subsystem":"skills"} {"source":"openclaw-managed","rootDir":"/Users/marcelo/.openclaw/skills","rootRealPath":"/Users/marcelo/.openclaw/skills","path":"/Users/marcelo/.openclaw/skills/video-processing-editing","realPath":"/Users/marcelo/.agents/skills/video-processing-editing","reason":"symlink-escape"} Skipping escaped skill path outside its configured root.
19:35:37+00:00 warn skills {"subsystem":"skills"} {"source":"openclaw-managed","rootDir":"/Users/marcelo/.openclaw/skills","rootRealPath":"/Users/marcelo/.openclaw/skills","path":"/Users/marcelo/.openclaw/skills/whisper","realPath":"/Users/marcelo/.agents/skills/whisper","reason":"symlink-escape"} Skipping escaped skill path outside its configured root.
19:35:37+00:00 warn skills {"subsystem":"skills"} {"source":"openclaw-managed","rootDir":"/Users/marcelo/.openclaw/skills","rootRealPath":"/Users/marcelo/.openclaw/skills","path":"/Users/marcelo/.openclaw/skills/youtube-transcribe-skill","realPath":"/Users/marcelo/.agents/skills/youtube-transcribe-skill","reason":"symlink-escape"} Skipping escaped skill path outside its configured root.
Log tail truncated (increase --max-bytes).

lethal ferryBOT May 12, 2026, 7:38 PM

#

Those symlink-escape skill warnings are probably not the cause of the terminated response. They just mean OpenClaw skipped skills in ~/.openclaw/skills that point over to ~/.agents/skills.

For the actual failure, the useful log lines are getting cut off. Re-run with a bigger tail:

openclaw logs --max-bytes 200000

Or live while reproducing:

openclaw logs --follow --max-bytes 200000

Then paste the part around the message that returns terminated, especially any lines containing failureKind, All models failed, CLI exceeded timeout, no output, provider name, or HTTP status.

If you want to clean up the skill warnings later, either remove those symlinks from ~/.openclaw/skills, or configure the real skill root / allowed symlink target. Docs section: https://docs.openclaw.ai/gateway/troubleshooting#skill-symlink-skipped-as-path-escape

trim delta May 12, 2026, 7:39 PM

#

19:39:00+00:00 warn agent/embedded {"subsystem":"agent/embedded"} {"event":"embedded_run_failover_decision","tags":["error_handling","failover","assistant","surface_error"],"runId":"active-memory-mp31ab9m-367fc313","stage":"assistant","decision":"surface_error","failoverReason":"timeout","profileFailureReason":"timeout","provider":"openai-codex","model":"gpt-5.2","sourceProvider":"openai-codex","sourceModel":"gpt-5.2","profileId":"sha256:46c4e4b775b6","fallbackConfigured":true,"timedOut":true,"aborted":true} embedded run failover decision
19:39:00+00:00 info plugins {"subsystem":"plugins"} active-memory: agent=main session=agent:main:telegram:direct:2102310958 activeProvider=openai-codex activeModel=gpt-5.2 done status=timeout elapsedMs=12291 summaryChars=0
19:39:05+00:00 warn agent/embedded {"subsystem":"agent/embedded"} {"event":"embedded_run_agent_end","tags":["error_handling","lifecycle","agent_end","assistant_error"],"runId":"4bd92bdd-62d9-41c9-97d0-b2828076b226","isError":true,"error":"terminated","failoverReason":null,"model":"gpt-5.4","provider":"openai-codex","rawErrorPreview":"terminated","rawErrorHash":"sha256:e8c95a2a7271","providerRuntimeFailureKind":"unknown"} embedded run agent end
19:39:13+00:00 warn agent/embedded {"subsystem":"agent/embedded"} {"event":"embedded_run_agent_end","tags":["error_handling","lifecycle","agent_end","assistant_error"],"runId":"4bd92bdd-62d9-41c9-97d0-b2828076b226","isError":true,"error":"terminated","failoverReason":null,"model":"gpt-5.4","provider":"openai-codex","rawErrorPreview":"terminated","rawErrorHash":"sha256:e8c95a2a7271","providerRuntimeFailureKind":"unknown"} embedded run agent end
1

lethal ferryBOT May 12, 2026, 7:40 PM

#

Yep, this points away from the skill warnings.

You’ve got two separate failures in that snippet:

active-memory ... status=timeout elapsedMs=12291 — the hidden memory recall is timing out after ~12s.
Main assistant call via openai-codex is ending with raw terminated on gpt-5.4.

Fast isolation test:

/active-memory off --global

Then send a simple message again. If replies start working, Active Memory is the trigger and you can re-enable it later with a higher timeout / safer model. Docs: https://docs.openclaw.ai/concepts/active-memory

If it still returns terminated, the main issue is your openai-codex model route. Please run:

#

openclaw models status
openclaw config get agents.defaults.models
openclaw config get agents.defaults.model

Also try switching the current session to a known-working configured provider/model with /model from chat. I don’t want to guess your available models, but the logs show openai-codex is the thing terminating, so testing another configured provider is the cleanest next cut. 🦞

trim delta May 12, 2026, 7:41 PM

#

marcelo@Marcelos-iMac ~ % /active-memory off --global
zsh: no such file or directory: /active-memory
marcelo@Marcelos-iMac ~ %

lethal ferryBOT May 12, 2026, 7:42 PM

#

Ah — sorry, that one is a chat slash command, not a macOS terminal command.

Send this to your OpenClaw bot in the same chat where you’re getting terminated:

/active-memory off --global

Not in zsh.

If you want to do it from terminal instead, use config. First inspect the current value:

openclaw config get plugins.entries.active-memory

The chat command is safer because it writes the right config key and keeps the command available for turning it back on later. Docs: https://docs.openclaw.ai/concepts/active-memory#commands

trim delta May 12, 2026, 7:42 PM

#

done

lethal ferryBOT May 12, 2026, 7:42 PM

#

Nice. Now send OpenClaw a normal tiny test message, like:

say only: ok

If that works, Active Memory was the thing pushing the run into failure. Leave it off for now, then later we can tune it with a bigger timeout / different recall model.

If it still replies terminated, run these and paste the output:

openclaw models status
openclaw config get agents.defaults.models
openclaw config get agents.defaults.model

That’ll confirm whether the remaining issue is the openai-codex route shown in your logs.

trim delta May 12, 2026, 7:47 PM

#

it replied now

#I am getting terminated as a response constantly