Run an agent manually
You don't need a flow to put an agent to work. You can start a run yourself from the task or epic you're looking at, or from the Agents page. Manual runs are useful for one-off work and for testing an agent before you automate it.
Run an agent on a task
Open the task and select the Delivery tab.
Under Agent, choose the agent you want to run. Only agents available for this task's team are listed.
If the task has a repository, choose how the agent should deliver its work, if the option appears.
Select Run agent.
The run and its progress appear on the same tab. If the button is unavailable, hover over it to see why, for example because a run is already in progress.
Run an agent on an epic
Open the epic.
In the Agent runs section, choose an agent next to Agent. Helpin suggests an epic planning agent when one is available.
Select Run.
Only one run can be active on an epic at a time. Past runs are listed in the same section.
Run an agent from the Agents page
Use the Agents page when the work isn't tied to the page you're on, for example a repository audit or a workspace-wide review.
Go to Automation → Agents and open the agent.
Select Run now.
In Run agent now, choose the Target. You can only choose targets that the agent is allowed to work on.
Fill in the target details, such as the Repository and Base branch, or the record to work on.
Optionally, add Run instructions for what this run should focus on.
Select Run now.
Follow a manual run
Manual runs work the same way as runs that flows start. They can pause for your approval, and you can follow them on the task or epic, or in Automation → Activity.
Workspace administrators can see every manual entry point on the Manual tab of Automation → Triggers.
Related
Was this article helpful?