T3 Code is an open-source agent console. It sells no models and charges no subscription; its job is to let you remotely control coding agents already installed on your own computer, from your phone, the web, or a desktop app. If you often think of a code change while you are out, or want to watch an agent work from the sofa, it sounds like a good fit — but the project is still very early, and whether it is worth the hassle depends on whether you can live with its rough edges.
What T3 Code does
T3 Code is positioned as an agent console. It does not provide a large model itself; instead, it connects to coding agents already installed and signed in on your computer, and lets you operate them remotely through a mobile app, a web app, and a desktop app. It supports seven agents in total — Claude Code, Codex, Cursor, Grok Build, OpenCode, Google Antigravity, and fx — provided those agents are already installed and signed in on your computer. In other words, it solves the problem of remote control; it does not give you a new coding assistant.
Repository details and setup cost
The project is hosted on GitHub, under the organization pingdotgg, with the project name T3 Code and the repository pingdotgg/t3code, under the MIT license. Installation is cheap: a one-click command-line install followed by running t3 starts the server, and you can also try it once with npx t3@latest. The desktop app also supports winget, Homebrew, .deb, AUR, and Nix. But low cost does not mean zero prerequisites: you must first install and sign in to at least one agent command-line tool — for example, Codex needs to be signed in first, and Claude Code needs its authentication completed — otherwise, once T3 Code is installed, there is nothing for it to control.
Three pitfalls to see clearly first
First, the project itself admits it is very early and that bugs should be expected, so be mentally prepared for instability. Second, the project currently accepts almost no outside contributions, and large feature submissions from outside are not accepted either, so if something breaks, the fix mainly depends on the maintainers, with limited help from the community. Third, it is only a console: the agents' capabilities, sign-in status, and costs are all decided by your own setup, and on a computer where no agent is installed, it has nothing to do. People who can accept these three points and genuinely need remote control can give it a try; people who are simply looking for a coding assistant that works out of the box need not spend extra time on it.
