History and run detail
The runs table, and the detail page with Input, Output, Files and every step.
The History tab
Every run of the agent: Started · Status · Trigger · Credits · Open.
| Column | Values |
|---|---|
| Status | pending running awaiting_input succeeded failed stopped |
| Trigger | manual (Run tab) · api · schedule · webhook · widget (embedded chat) · comparison |
A run at awaiting_input is waiting for you — open it to approve or answer.
The detail page
Input — the request and attached images. Output — the final result; if the run failed, the RunFailure block points at the failing step. Files — files the agent produced (an xlsx, say), downloadable.
Steps — a table # · Step · Model · In · Out · Credits, one row per step of kind model,
retrieval or tool. Expand a row for Given (what the step received) and Produced (what
it returned) as raw JSON. This is where you find out why the agent answered the way it did.
Resume / Stop / Retry need agentRun.control.
Total credits across runs are also on Usage under the agent operation.
Versions — compare and go back
The Versions tab — the version list, a diff between two versions, restoring an old one, and trying one question against several versions.
Connected accounts
Connect Google, Slack, GitHub and Notion so agents can use Gmail, Drive, Calendar, Sheets, Slack channels, GitHub issues and Notion pages.