#Model timeout
1 messages · Page 1 of 1 (latest)
There are a few different timeouts in OpenClaw, and the right knob depends on what actually timed out:
- Overall agent run:
agents.defaults.timeoutSeconds, default is 172800s (48h) - LLM idle timeout:
agents.defaults.llm.idleTimeoutSeconds, this is usually the important one for slow models. Current source says 120s if unset, but some older releases used 60s - Shell command / exec timeout:
tools.exec.timeoutSec, default is 1800s (30m)
Useful examples:
openclaw config set agents.defaults.llm.idleTimeoutSeconds 300
openclaw config set agents.defaults.timeoutSeconds 3600
openclaw config set tools.exec.timeoutSec 7200
openclaw gateway restart
If it is just one exec call, you can also raise the timeout just for that command instead of changing the global config.
Docs:
- https://docs.openclaw.ai/concepts/agent-loop
- https://docs.openclaw.ai/gateway/background-process
- https://docs.openclaw.ai/cli/config
If you paste openclaw --version and the exact timeout error line, I can tell you which one you’re hitting.
2026.4.15 : Seems like I need to update idleTimeoutSeconds then
Yep, if the model request itself is going idle, agents.defaults.llm.idleTimeoutSeconds is the right knob.
I’d try something like:
openclaw config set agents.defaults.llm.idleTimeoutSeconds 300
openclaw gateway restart
A few notes:
300= 5 minutes0disables the idle timeout entirely- if you’re on a slow local model,
300or600is usually a reasonable starting point