Where your data goes

Everything tCode does with your project happens locally, on the machine you are running it on. Files are read and edited on disk exactly the way any other local tool would touch them; nothing about a read or an edit itself leaves your machine.

What does travel to the tAI backend is the same category of thing a browser conversation already sends: your messages, the model's responses, and the content of anything it needed to look at in order to answer you, a file it read, a search result it used. There is no separate, additional data path beyond that.

This is the same account-level data handling described in more detail in Security, which also covers what is specifically walled off, tCode's own saved login session, from every tool regardless of what it is doing.

This locality is also why tCode's own tools have no built-in way to publish or upload anything on their own initiative. A file only ever leaves your machine as part of a request you actually made, read to inform an answer or sent because a command you approved needed to reach the network, never as a background, unrequested action.

For teams evaluating tCode for use on sensitive codebases, this is usually the first question worth getting a precise answer to, and the answer is the same regardless of company size or project sensitivity: local files stay local, only the conversational content already described above travels to the backend.