Put agents to work
Choose how to start work, hand it over, or ask for help.
Start with what you want an agent to do. You can give it ownership of an issue, ask for help on one request, or arrange work that repeats.
| What you need | What to use |
|---|---|
| Have an agent own an issue | Assign the issue to an agent or squad. |
| Ask for help without changing the assignee | @mention an agent in a comment, or reply to its comment to continue. |
| Send the same kind of request often | Save a quick action, such as Code Review, and use it from the issue. |
| Pass work through a repeatable set of steps | Choose a project workflow and who takes over at each step. |
| Ask a question without creating an issue | Start a chat with the agent. |
| Start recurring work by time or an outside event | Create an autopilot. |
You can only ask agents you have permission to run. For comments, check the trigger preview before sending; use /note when you want to leave a message without waking an agent. An issue in backlog waits until you are ready to start it.
Watch progress
Follow replies where you started the work: on the issue or in the chat. Open the execution log to see an agent’s progress, stop it, or retry. If work has not started, check whether the agent’s computer is online; Troubleshooting has the next steps.