Skip to main content

Live terminal

OtoDock runs the real coding agents, Claude Code and Codex, as its engine. A chat normally presents their work as a clean streaming conversation. When you want the full experience, the same chat can run as an interactive terminal: the agent's native interface, live in your dashboard, with you in the loop.

The actual CLIs, not a lookalike​

Terminal mode opens the real thing. It is the same interface you would get by running Claude Code or Codex yourself, with everything OtoDock normally prepares already in place:

  • The agent's identity: its prompt, context files, and memory.
  • Tools and MCPs: every tool the agent is allowed to use, already connected.
  • Credentials: the AI engine account is injected for the session, so nothing asks you to log in.
  • Files and security: the agent's workspace, and the same sandbox and permission rules as every other session.

You watch the agent's interface exactly as it renders, plans, spinners, colors and all, and you can type into it at any moment.

Turning it on​

Terminal mode is available out of the box:

  • Per chat: flip the terminal toggle in the chat's toolbar. Each chat remembers its choice.
  • Per agent: set the agent's default execution mode in Agent Config, and every new chat with that agent starts in terminal mode.
  • Platform wide: admins can disable the feature entirely in System Settings ("Interactive terminal sessions"). Every session then runs in the standard view and the toggles disappear.

Type in it, or keep chatting​

Work directly in the terminal if you like the keyboard feel of the CLIs. Or keep using the normal message box under it: anything you send there lands in the terminal as if you had typed it, including dictated voice messages. Both paths reach the same session, so mix them freely.

Approvals, in the agent's own words​

In terminal mode the agent asks for permission the way the CLI natively does: the question appears inside the terminal and you answer it right there. Your chosen permission mode still applies, from approving each sensitive step to fully autonomous.

Switch views whenever it helps​

A chat is never locked to one mode. Switch between the terminal and the standard chat view at any time; the conversation carries over with full context in both directions. The transcript is saved like any other chat, so it shows up in your history, works with search, and can be resumed later, even after a restart.

Terminal tasks

This works for scheduled tasks too. Give an agent a terminal default and its task runs open as live terminals you can watch, and step into, while they work.

Good to know​

  • The terminal matches your dashboard theme.
  • Idle terminal sessions close on their own after a while. The conversation stays in the chat, and your next message picks it up where it left off.
  • Multi agent meetings always run in the standard chat view; a single terminal cannot host several agents at once.

Next steps​