#stuck
2 messages · Page 1 of 1 (latest)
Not usually one bug, sadly. Most “agent is stuck” cases fall into a few buckets:
-
Model/tool-calling is weak
Biggest culprit, especially with local/Ollama/OpenAI-compatible backends. They can stall, ignore tool results, or loop. -
Gateway/auth/pairing issue
The UI/channel is connected enough to show activity, but the actual run never starts or loses auth. -
Bad session/bootstrap state
A stale session, stuck bootstrapping state, or leftover workspace state can wedge runs. -
Tool polling loops
Agent keeps calling the same tool with no progress. OpenClaw has loop detection for this, but it’s off by default unless you enabled it.
First thing, paste these: