The Runs panel is the project execution history for agents, tasks, and workflows. The list shows run context, status, and timing. Filter by run ID or status, sort the list, load more records, or refresh it after starting work elsewhere.
Create a run
Create Run offers three run kinds:
An invalid JSON object or missing required target prevents submission. Task and workflow actions in their respective panels can open this form with the target selected.
Inspect a run
Open a row to inspect its subject, status, timestamps, output, failure details, and event history. Events can be paginated and exported as CSV. An agent run linked to a conversation offers Open chat.
If the event list is empty, refresh it. The run may not have emitted events yet.
Run events show model calls and tool calls. Use Logs for runtime output. The run guide describes execution semantics.