FRIDAY, NOVEMBER 20, 2026 Tech news, minus the noise.
N News Cast
LATEST
Home  /  Dev
Dev

T3 Code Put Every Coding Agent in One Window and Developers Noticed

T3 Code, an open-source control surface for Claude Code, Codex, Cursor, and other coding agents, gained more than 14,000 GitHub stars in a single week. It does not replace your agents. It gives them one shared window on desktop, web, and phone.

T3 Code Put Every Coding Agent in One Window and Developers Noticed
Source: Markus Spiske via Unsplash (CC0)

The fastest-growing AI project on GitHub is not a model

T3 Code is the fastest-growing AI project on GitHub this month, and it is not a model, not a coding agent, and not a startup with a funding announcement. It is an open-source control surface for the coding agents you already use. The project, from the ping.gg team behind Theo's T3 stack, puts Claude Code, Codex, Cursor, OpenCode, Grok Build, Google Antigravity, and others behind one shared interface on desktop, web, and mobile. In the week of November 2 to 8, it gained 14,125 stars. By November 15, the repository had passed 60,000 stars, with about 11,000 forks. It started February 2026 with no audience at all.

The idea is simple enough to explain in one sentence: if you pay for several coding agents, you should not need several windows, several CLIs, and several mental models to use them. T3 Code drives the agents you have already installed and signed in to, and gives them a single console with per-task Git workflows and remote access from your phone. It ships no model of its own and it does not resell tokens. Your provider bills stay exactly where they are.

What T3 Code actually does

The distinction matters, so it is worth being precise. T3 Code calls itself an agent harness control surface, and the README is narrower still. It does not write code itself. It remote-controls the coding agents on your machine through their own CLIs and APIs, and organises their work into threads with Git state attached. Start a task on your desktop, check on it from your phone on the train, and review the diff when you get home. The server runs locally, the clients connect to it, and nothing about your code leaves your machine except through the providers you already use.

Installation is straightforward by the standards of this kind of tool. The CLI installs from the official release, the server listens on a local port, and you pair your phone through T3 Connect. The project was at version 0.1.7 on November 4 and 0.1.8 on November 9, which tells you how fast it is moving and how early it still is. The README is blunt about the state of things: the team says it is very early in the project and to expect bugs. That honesty is part of why developers trust it.

The supported list is wider than the headline names suggest. The homepage names Claude Code, Codex, Antigravity, OpenCode, Cursor, and Grok. The installation guide lists ten providers, including Kiro, Pi, Muse Code, and fx. The point is not to lock you into one agent. It is the opposite: to make switching between agents cheap enough that you use the right one for each task instead of the one whose window happens to be open.

Why developers are piling in

The star count is a signal, not a verdict, but the size of it is hard to ignore. Fourteen thousand stars in a week puts T3 Code ahead of every other AI project on GitHub that week, and the growth has continued since. The reason is not hype. It is that the problem is real and widely felt. Developers who adopted coding agents early now juggle several subscriptions, and each one has its own interface, its own way of handling context, and its own idea of where your code lives. T3 Code is the first tool to treat that fragmentation as the problem to solve, rather than as a reason to pick a winner.

It also helps that the project is MIT licensed and free. There is no token markup, no seat pricing, and no vendor lock-in beyond the providers you already pay. The project's homepage claims more than 500,000 developers use T3 Code, a figure nobody has independently verified, so treat it as marketing rather than fact. The GitHub numbers, which are public and checkable, are impressive enough on their own.

What to watch before you install it

Enthusiasm is warranted, but so is caution. T3 Code is nine months old and shipping weekly releases with breaking changes. The open issue count is in the thousands, which is normal for a project growing this fast and still a sign that you should not bet a production workflow on it yet. Security-minded developers should also think carefully about what a control surface means: a tool that can drive your coding agents from your phone is a tool with deep access to your repositories, and the security model of that remote access deserves a careful read before you enable it.

The comparison worth making is with the last generation of developer tools. GitHub rebuilt its Git storage this month to handle 5,000 AI agents pushing at once, and Cloudflare open-sourced Kitesurf, a browser engine built for agents. The whole stack is being rebuilt around the assumption that agents, not humans, are the primary users of developer tools. T3 Code is the interface layer of that rebuild: the place where a human watches, steers, and reviews what the agents are doing.

If you use one coding agent and you are happy with it, T3 Code solves a problem you do not have. If you use two or more, or you want to check on a long-running agent task from your phone, it is worth an afternoon. Install it, wire up one provider, run a real task through it, and see whether the single window is worth the early-stage rough edges. For a lot of developers this month, the answer has been yes, and the star count is the receipt.

AS
News Cast Editorial Team

We cover AI, gadgets, startups and security — original reporting and explainers, written by humans, for humans. Got a tip? Reach us via the contact page.