FAQ

Is tCode free?

tCode itself is free to install and run. It uses your existing tAI account, so usage follows whatever plan that account is already on, exactly as if you were using tAI in the browser.

What can it do without asking me first?

Reading a file, searching your code, and searching the web all run without interrupting you, since none of them can change anything. Writing a file, editing one, or running a shell command always shows you exactly what is about to happen and waits for a yes, unless you have already turned off that prompt for a specific tool with /config auto-approve.

Can I undo something it changed?

Yes. Every turn that changes files is checkpointed automatically beforehand, and /rewind restores your working tree to how it was before the most recent one, or further back with /rewind <n>. See Checkpoints and rewind.

How is this different from using tAI in the browser?

Same account, same models, same conversation history and memory. tCode adds direct access to your local filesystem and shell, so it can actually read, edit, and run your project instead of you pasting code back and forth between a browser tab and your editor.

What platforms does it support today?

Windows and Linux, on Python 3.9 through 3.14. See Installation and download.

Does it need its own API key?

No. tCode signs in with your normal tAI account. API keys from console.artfical.com are for a separate use case, calling tAI models from your own code, not for using tCode itself. See Accounts and API access.

Can more than one person use the same paired machine?

A machine is paired to a single tAI account through /pair. If a different account needs remote control of the same machine, it needs to be paired separately from a Remote Control chat under that account.

Does tCode work offline?

No. Every turn talks to the tAI backend the same way a browser conversation would, so a working network connection is required. Purely local actions inside a single tool call, reading a file, running a command, still happen on your machine, but the reasoning that decides what to do next always requires reaching tAI.

Can I use tCode in a script or a CI pipeline?

tCode is built as an interactive terminal tool, centered on the permission prompts described in Permissions and safety, rather than a headless automation tool meant to run unattended. It is not currently designed for a fully non-interactive pipeline where nothing is available to answer a permission prompt.

Does it work the same way in every project, or does it need per-project setup?

No project-specific setup is required beyond the one-time folder-trust confirmation the first time you run tCode there. It reads whatever is actually in the project to understand conventions and structure, rather than depending on a configuration file you would need to write and maintain yourself.