Autoresearch
Autoresearch runs a saved research runbook on a schedule or from a webhook. Use it for recurring tasks such as a literature watch after you have tested the same work manually. You need a workspace, an appropriate assignee and a machine that will be available when the trigger fires.
Create a repeatable task
- Open Autoresearch and choose New autoresearch.
- Enter a clear name and a runbook with the goal, inputs, ordered steps and expected output. The agent reads this runbook on every run.
- Select the assignee and review its stage scope.
- Choose Create a study to put each run on the board, or Run only to report back without creating a study.
- Choose Schedule or Webhook. For a schedule, review the time, recurrence and timezone shown by the product.
- Create the autoresearch and inspect its Triggers section.
Check that a trigger actually exists. Creation can succeed while trigger creation is refused; the app reports that separately. An entry with no triggers can still be run by hand.
Test and inspect
Use Run now on an active autoresearch for a controlled first attempt. Inspect Runs, then open the created study or run details. Confirm the output matches the runbook and note any source or execution failure. See runs and studies.
Control future work
Use Pause to stop future triggering and Resume when the task is ready again. Read the displayed status and any pause reason. Check execution results after changes rather than assuming the next schedule will succeed.
A webhook URL is shown only when freshly created. Copy it securely if you need it and treat it as a trigger secret; do not put it in a study or public screenshot. Each POST to that URL can start work.
Success means you can identify the saved runbook, its actual trigger and the output of a test run. If execution fails, inspect the machine and connector requirements before relying on unattended work.