T3 Code

T3 Code is a desktop application that runs several AI coding agents from one interface. The agents it drives are existing command-line tools and services: Claude Code, Codex CLI, OpenCode, Cursor, Grok CLI and Google Antigravity. T3 Code does not provide its own model. You sign in with the accounts you already have, and the app orchestrates those agents, shows their work and manages the Git side of what they produce. It runs on macOS, Windows and Linux, with iOS and Android apps that connect back to your machine.

The central design choice is that T3 Code does not resell model access. There are no token bundles to buy and no quota caps added on top of your subscriptions, so usage limits stay exactly as they are in Claude, OpenAI, Cursor or Google. You can switch models mid-thread instead of starting a new conversation, and the project says more agent harnesses ship each week. Antigravity connects through a Google sign-in, while the others use their own login commands. Installation is a single download for desktop, a shell script or a one-off npx command for the terminal, and the mobile apps pair with the desktop instance over your own network.

Git and pull requests are built into the workflow. Every agent thread writes to its own branch, so parallel experiments remain separate until you decide to keep them. When a change looks right, one button commits, pushes and opens a pull request on GitHub with a generated title, body and changelog, and you can review the diff inline before it leaves your machine. Draft pull requests, stacked changes and amend flows are all supported, and the app reuses your existing GitHub authentication.

The application is open source under the MIT licence, published by T3 Tools Inc as github.com/pingdotgg/t3code. The marketing page invites forks outright: you can restyle every surface, wire in your own tools and models, self-host the build or distribute your own version. Releases are numbered in the 0.0.x range and the project ships quickly, so the interface and the roster of supported agents change regularly rather than once a year.

Practically, that makes T3 Code an orchestration layer rather than a replacement for the agents themselves. It adds a shared surface, branch isolation, diff review, generated pull requests and remote access from a phone. Nothing about your existing command-line setup changes when you install it, and because the code is public you can inspect or extend anything that does not suit you. It does assume you already pay for at least one agent and have a development environment worth connecting. Support and design discussion happen on the project's Discord server rather than through a paid support desk.