Skip to content

Messages, tools and hooks ​

History explains where an answer came from and which stage failed.

ItemInspect
User messageTask text and supplied context
Agent messageProgress explanation or final answer
Tool callArguments, command, result and exit code
HookData preparation or result publication
Technical journalStates and operational events

Tool inputs appear before the call completes. Expand the output separately. Large messages and results may need expansion. Truncated output is marked explicitly.

If the agent produced an answer but delivery failed, inspect after_run. If the agent never started, inspect workspace preparation and before_run.

History loads in pages. For long runs, load earlier entries or use the history API. Markdown images appear as links so viewing history does not make background requests to third-party servers.

Troubleshooting · History API