Browse Workspace
Workspace · Reuse and Automate

Scheduled Tasks

Configure recurring agent work with explicit skills, connections, permissions, timing, results, and run history.

2 min readLast reviewed July 27, 2026

#Build and test the task

Configure and test the complete task before relying on its unattended schedule.
Screenshot previewConfigure and test the complete task before relying on its unattended schedule.
Scheduled Task settings with timing and workspace configuration controls
  1. 01

    Prove the request manually

    Run the task in a workspace and verify a complete, representative result.

  2. 02

    Choose the schedule

    Set the frequency, time, and time zone shown in the task settings.

  3. 03

    Specify skills

    Choose the skills the agent may use instead of hiding a long repeatable method in the prompt.

  4. 04

    Specify connections

    Choose only the connections the unattended work needs.

  5. 05

    Verify permissions

    Confirm the effective tool access can complete without waiting for a person.

#Permissions for unattended work

A Scheduled Task cannot depend on someone being present to answer an approval prompt. A tool set to Ask may prevent the unattended run from completing the intended action.

Review the task’s chosen connection and skill access together with the applicable personal, workspace, and administrator policy. Use the narrowest access that still completes the verified task.

A Scheduled Task needs an explicit schedule and enough approved access to run without a person present.
Screenshot previewA Scheduled Task needs an explicit schedule and enough approved access to run without a person present.
Scheduled Task settings with a schedule and unattended workspace permissions

#Review run history

Use run history to confirm the schedule ran and the result was actually delivered.
Screenshot previewUse run history to confirm the schedule ran and the result was actually delivered.
Scheduled Task history showing completed and previous recurring runs
  • Check whether each run started and finished at the expected time.
  • Open the result or artifact and verify freshness and completeness.
  • Inspect connection or permission failures before changing the prompt.
  • Pause the schedule when the source, access, or expected output is no longer valid.