The agent is reporting that the system is completely frozen. Every command is timing out. It can’t run anything until it recovers. Every time I send a message it seems to cost a lot of money just to send me back a response.
#Exec system is completely frozen
coral saddle
cloud nymphBOT
halcyon cragBOT
Stop sending messages to the agent for a minute — if tools are timing out, each “try again” can burn tokens without making progress.
To debug “exec/system frozen”, we need to separate gateway health vs tool/node health. Can you paste the output of these (in order) as code blocks?
openclaw --versionopenclaw status --allopenclaw gateway probeopenclaw gateway status(looking for RPC probe: ok)openclaw doctoropenclaw channels status --probe- If you use nodes for exec:
openclaw nodes status openclaw logs --follow(grab ~30 lines around the timeouts)
Docs ladder (same checklist): https://docs.openclaw.ai/help/troubleshooting and deep runbook: https://docs.openclaw.ai/gateway/troubleshooting
Immediate “get unstuck” actions that are usually safe:
- If
gateway statusis unhealthy or CLI calls hang:openclaw gateway restart
- If you’re on a VPS / long-running host and it’s still wedged after restart: a full machine reboot has a surprisingly high success rate for clearing stuck state.
On the cost issue: once we have it responding again, set up a cheap “fast chat” agent and keep the expensive model for heavy work (docs section):
https://docs.openclaw.ai/help/faq#can-i-run-a-fast-chat-agent-and-an-opus-for-coding-agent
Send the outputs and I’ll tell you exactly which branch you’re in (gateway unreachable vs channels/pairing vs node/exec stuck).