The terminal in Oda

The side panel next to your chat can hold a terminal of your own: a shell that starts in the chat's workspace folder, where you type commands yourself. It stays yours: Oda does not type in it, does not read its output, and nothing from it joins the chat.

Open a terminal

1. In the side panel beside your chat, choose New terminal; its card reads Run commands in the Session workspace. A terminal opens in its own tab, running the shell you chose last, or the environment's default the first time.
2. To use another shell, open the arrow beside New terminal (named Choose shell) and pick one from the list. Your pick opens right away and is remembered for later terminals.
3. To open a second terminal, use New tab in the panel and choose New terminal again. Each terminal gets its own tab, and one tab holds one terminal.

A new terminal starts at 80 columns by 24 rows, smaller when the environment caps it lower. The terminal comes from the Terminal tab switch under Settings > Developer > Advanced features; if the entry does not appear, check that switch first. Advanced features covers the switches.

Choose your shell

• The list shows the environment's default shell first, then the shells verified on this computer: zsh, bash, fish, pwsh, powershell and cmd. Where the environment names no default, the fallback is /bin/sh on POSIX systems or cmd.exe on Windows.
• Only shells Oda can verify are offered, and a shell path it cannot verify is refused: Selected shell is not available in this execution environment. Oda does not start a different shell in its place.
• Your last choice is remembered on this computer, and the list marks it the next time you choose.

Rename and resize the tab

• Double-click the tab's title to rename the terminal; the editor's field is Terminal name. Press Enter to save the name, or Escape to cancel; a name is between 1 and 120 characters long.
• The screen refits as you resize the panel or the window, and it never grows past the size cap in the table below.

What the tab shows

• While a terminal opens, its tab reads Reading terminal environment…, then Starting…, then Connecting…; the shell's screen appears once it is attached.
• When the shell exits, the tab keeps the final output and shows Process exited (<code>) with the exit code. The shell is never restarted on its own; use New terminal in the tab for a fresh one.
• A lost connection keeps the screen as it was and offers Reconnect; the shell keeps running either way.
• A terminal that cannot start reads Unavailable and offers Retry.
• When another view of the same terminal owns the keyboard, this view reads This view is read-only. and offers Take control; the shell keeps running.

Keep a terminal running

• A terminal keeps running while it is not visible: collapsing the panel, switching tabs or chats, and floating or full-screen views leave it running. A busy command is not stopped by a timer.
• Reloading Oda reattaches each terminal tab to its shell and restores the screen. If the shell is gone, for example because Oda quit in between, the tab reads This terminal no longer exists. Open a new terminal. and a New terminal button replaces it in place. Nothing restarts on its own.
• A terminal that no open tab holds is closed after about two hours without activity. Only time in which the shell is definitely idle counts, so a command that keeps working is not reclaimed by this timer.

Close a terminal

• Closing a tab removes it at once and ends the shell and every process it started, in the background. Detaching a view or hiding the tab never ends the shell.
• A closed terminal cannot be reopened, and Oda refuses to recreate it; open a new terminal when you need one.
• If the cleanup cannot finish in one go, the app remembers the request and retries it; you do not need to do anything.
• An exited terminal stays visible and keeps counting toward the chat's terminal limit until you close it.

Your shell, not the sandbox

Commands you type here run with the rights of your own account on this computer, exactly as if you opened the shell yourself. Oda does not elevate them, and the chat's access mode, its approval prompts and Oda's sandbox do not apply to this terminal; changing the access mode leaves running terminals as they are. See Workspaces, files and permissions for the modes themselves.

The shells Oda keeps for its own commands are a different surface; see Persistent shells.

Limits

Questions, answered

Where does a new terminal start?

In the chat's workspace folder. The terminal belongs to the chat, and it opens there with the shell you chose last, or the environment's default the first time. Workspaces, files and permissions explains workspace folders.

Which shells can I choose from?

The environment's default comes first, then the shells verified on this computer: zsh, bash, fish, pwsh, powershell and cmd. Your last choice is remembered and marked in the list, so you usually just press New terminal.

My shell is not in the list. Can I still use it?

Only shells Oda can verify are offered. A shell path that cannot be verified is refused with Selected shell is not available in this execution environment, and Oda does not start a different shell in its place. Pick one of the offered shells, or reinstall your shell on this computer and open the list again.

Does Oda see what I type in the terminal?

No. The terminal belongs to you: Oda does not type in it and does not read its output, and nothing typed or printed here joins the chat's transcript. It works like a shell you open yourself, next to your chat.

Is the terminal sandboxed like the commands Oda runs?

No. Commands you type run with the rights of your own account on this computer, and Oda does not elevate them. The chat's access mode, its approval prompts and Oda's sandbox do not apply to your terminal, and changing the access mode leaves running terminals as they are.

What happens to my terminals when Oda reloads?

Each terminal tab reattaches to its shell and restores the screen. If the shell is gone, for example because Oda quit in between, the tab reads This terminal no longer exists. Open a new terminal. and a New terminal button replaces it in place; nothing restarts on its own.

Can I reopen a terminal I closed?

No. Closing ends the shell and every process it started, and a closed terminal cannot be recreated; open a new terminal instead. If the cleanup cannot finish in one go, the app retries it in the background.

The same terminal is open in two windows and one will not accept typing. Why?

Only one view of a terminal owns the keyboard at a time: the newest one. The other view reads This view is read-only. and offers Take control; choose it in the view you are working in. Both views see the same shell, and the shell keeps running.

Why did my terminal close while I was not using it?

A terminal that no open tab holds is closed after about two hours without activity. Only time in which the shell is definitely idle counts, so a command that keeps working is not reclaimed. Open a new terminal when you need one again.

Pasting a lot of text stops the terminal. Why?

Typing is buffered up to 64 KiB at a time, and a very large paste can fill that buffer: the tab shows The input buffer is full. Reconnect and try again. Use Reconnect and send the text in smaller pieces; the shell keeps running.

A command printed a flood of text and the tab stopped updating. What should I do?

That view fell behind the terminal's output buffer. It reports Terminal output consumer exceeded its buffer; reconnect to recover the current screen; reconnect, and the current screen is restored. The shell keeps running while you do.