Project terminal

Open a project terminal#

Choose View → Toggle terminal, or use the desktop Toggle terminal control, to open the Project terminal below the Code canvas. The terminal is scoped to the current project. New terminal creates another session, with up to twelve project terminal sessions available. A session reports Starting…, Ready, Exited, or Unavailable so the current state remains visible.

The panel can be resized with the Resize terminal panel handle. Stop ends the selected session, Restart starts it again when the project environment permits, and the close action removes the session from the terminal rail. Terminal sessions are useful for commands, logs, and local checks while the task transcript and review surfaces remain open above.

Separate setup from project work#

View → Runtime setup opens the user-owned runtime setup terminal. This surface is for installing or authenticating local runtimes and is separate from an agent transcript and from the project terminal. Runtime setup directs password, MFA, and key entry to the vendor’s browser flow. The Settings → Runtimes & Models page provides the corresponding Install, authenticate, Update, and Check all controls.

Keeping setup separate prevents a vendor credential prompt from becoming task content. Code also keeps terminal commands within the selected project and permission boundary. The active task’s permission mode still determines whether an agent may perform a command or must request approval.

Desktop and mobile scope#

Project terminal is a desktop Code surface. The phone’s Remote task view can supervise a paired run, inspect work surfaces, and open the remote desktop browser, but it does not provide a free local project terminal. Account-backed Chat has its own conversational surface and does not gain a Code project terminal through the Chat account. Runtimes, Tasks, Remote access, and Troubleshooting describe the related boundaries.