Terminal

Run commands in an interactive shell next to your chat. Each terminal is a real shell session that starts in the workspace's root folder, so commands run against the same files the agent is working with.

Working with the agent

The terminal is yours to drive, but it's connected to the chat:

• The agent can open a terminal for you, starting in the workspace folder, for you to work in. It won't type or run commands in it (the agent runs its own commands through a separate, permission-gated shell).
• The agent can read what your terminals print, such as a dev server or test watcher you're running, to help you spot issues and offer assistance.

Work in multiple terminals

Open more than one terminal at a time — each is numbered and runs independently, so a command in one doesn't affect another. The tab title updates to show the command that's currently running.

Sessions that survive reloads

A terminal keeps its scrollback and stays running across window reloads. If the view reloads, it reconnects to the same session rather than starting over, so a long-running command keeps going. Closing a terminal, though, ends its session and stops whatever it's running.