What is T3 Code?
T3 Code, from Theo Browne's ping.gg (T3 Tools), calls itself an agent harness control surface. It does not include a model or an agent of its own; it drives the coding-agent CLIs already installed and signed in on your computer, Claude Code, Codex, Cursor CLI, Grok Build, OpenCode and Google Antigravity, and exposes them through a mobile app for iOS and Android, a web app at app.t3.codes and an Electron desktop app.
The pitch is remote-ready and open: run the t3 server on your machine and reach your agents from a phone while they run at home, with the project licensed MIT so you can fork it. The README is blunt that there is nothing to sell, citing Codex's desktop app, Conductor and Cursor Glass as inspirations that did not meet its bar.
It installs with a shell script, npx t3@latest, or packages for winget, scoop, Homebrew, apt, the AUR and nixpkgs. The repo has about 32,600 stars and 8,600 forks; the latest release is v0.1.5 (29 October 2026).
Is T3 Code worth it?
Most agent front-ends tie you to one model or one vendor's app. T3 Code does the opposite: it is a thin control layer over whichever agent CLIs you have authenticated locally, so your existing Claude Code, Codex or Cursor subscription does the work and T3 Code just gives it a better surface, including a real mobile app that lets you kick off and watch runs from a phone. For anyone who already lives in several agents, that neutrality is the draw.
The openness is unusually load-bearing. The README states there is nothing to sell and that if the project ever goes the wrong way, you should have everything needed to fork and build the editor you want, and the MIT licence backs that up. It reads as a deliberate contrast to the closed desktop apps it names as inspiration, and with about 8,600 forks against 32,600 stars, people are taking the fork invitation more than most repos see.
The cost of being a control surface is that it inherits every dependency underneath it. You must install and log in to at least one provider CLI first, and the agents run with those tools' own permissions on your machine; T3 Code adds reach, not a sandbox. Running the server and opening it to a phone is a remote-access path worth securing deliberately, which the project's remote-ready framing makes easy to underestimate.
Maturity is the main caveat. Despite the star count the current release is v0.1.5, and the tracker carries roughly 4,000 open issues, the churn of a popular project shipping fast across six agent back ends, three app platforms and a long list of package managers. It is the most capable open multi-agent control surface available, and also one changing week to week, so pin a version if you depend on it.
Where it earns its place
A developer who pays for Claude Code and Codex and works across a laptop and phone
Start and monitor agent runs on the home machine while away from the desk
Install the provider CLIs and sign in, run the t3 server with t3 service install, and open the mobile or web app to drive them. Free software on top of the subscriptions already paid for; secure the server before exposing it beyond localhost.
A small team comparing agents on the same task
Run Claude Code, Codex and Cursor side by side without three separate UIs
Authenticate each CLI once; T3 Code surfaces them through one control surface so a developer switches back ends without leaving the app. The limit is that each agent still bills against its own plan.
An engineer wary of vendor lock-in in agent tooling
Adopt a front-end they can keep even if the project changes direction
The MIT licence and large fork count mean the editor can be forked and self-built; the trade-off is depending on a v0.1.5 project with heavy issue churn, so pin a release for stability.
Before you commit
- Setup
- curl -fsSL https://t3.codes/install.sh | sh (or npx t3@latest to try), then run t3 to start the server and open the local web app; t3 service install keeps it running. Desktop app via GitHub Releases, winget, scoop, Homebrew, apt, AUR or nixpkgs. You must install and log in to at least one agent CLI first.
- Integrations
- Drives Claude Code, OpenAI Codex, Cursor CLI, Grok Build, OpenCode and Google Antigravity; clients are iOS and Android apps, a web app at app.t3.codes and an Electron desktop app.
- Data and privacy
- Runs on your own machine against your local agent CLIs; those agents send code to their own model providers under their own terms. Exposing the server for remote use is a network-access decision you own. MIT licence; no security advisories published.
- Cost at scale
- The software is free; cost is the agent subscriptions it drives (Claude Code, Codex, Cursor and so on), billed per their own plans. No hosted fee from T3 Code itself.
- Limits
- v0.1.5 and roughly 4,000 open issues; adds no sandboxing over the agents' own permissions; depends entirely on the provider CLIs being installed and authenticated; fast release cadence.
What’s good
- Drives Claude Code, Codex, Cursor, Grok Build, OpenCode and Antigravity
- Real iOS and Android apps, plus web and desktop
- MIT-licensed and explicitly fork-friendly
- Uses the agent subscriptions you already have
What’s not
- No model or agent of its own; you supply the providers
- v0.1.5 with roughly 4,000 open issues
- Remote access to local agents needs securing yourself
- Agents run with their CLIs' full machine permissions
Best for
Developers who run several coding agents and want one open surface, including a phone, to drive and watch them.
T3 Code alternatives
T3 Code: common questions
- Is T3 Code worth it?
- T3 Code is the remote control for the coding agents you already pay for, and the one that lets you walk away from your desk while they work. We rate it 4.2 out of 5.
- How much does T3 Code cost?
- Free and open source (MIT). You bring your own agent subscriptions (Claude Code, Codex, Cursor and others).
- What are the downsides of T3 Code?
- No model or agent of its own; you supply the providers; v0.1.5 with roughly 4,000 open issues; remote access to local agents needs securing yourself; agents run with their CLIs' full machine permissions.
- Who is T3 Code best for?
- Developers who run several coding agents and want one open surface, including a phone, to drive and watch them.
- How much does T3 Code cost for a team or in production?
- The software is free; cost is the agent subscriptions it drives (Claude Code, Codex, Cursor and so on), billed per their own plans. No hosted fee from T3 Code itself.
- Is T3 Code safe for company data?
- Runs on your own machine against your local agent CLIs; those agents send code to their own model providers under their own terms. Exposing the server for remote use is a network-access decision you own. MIT licence; no security advisories published.
- What are the best T3 Code alternatives?
- The closest alternatives we have reviewed are OpenCode, Claude Squad and Vibe Kanban.
Answers checked Oct 30, 2026.
Make T3 Code? Show this rating on your site
