Autopilot is currently an experimental feature that lets the agent keep working autonomously until a task is completed, rather than pausing for your approval at each step.
How to enable it:
- First, activate experimental mode — either launch with copilot --experimental or run /experimental from within the CLI (this persists, so you only need to do it once).
- Then press Shift+Tab to cycle through modes — you'll see a new Autopilot mode alongside the existing Interactive and Plan modes.
In autopilot mode, when you submit a prompt, the agent will continue executing steps (file edits, commands, tool calls) without stopping for
confirmation until the task is done. It's ideal for well-defined tasks where you trust the agent to run end-to-end.
Autopilot is currently an experimental feature that lets the agent keep working autonomously until a task is completed, rather than pausing for your approval at each step.
How to enable it:
In autopilot mode, when you submit a prompt, the agent will continue executing steps (file edits, commands, tool calls) without stopping for
confirmation until the task is done. It's ideal for well-defined tasks where you trust the agent to run end-to-end.