AI product Open source

T3 Code

T3 Code is an open-source agent-harness control surface for coding agents running on a developer's computer. It provides iOS and Android mobile apps, a web app, and an Electron-based desktop app for controlling locally configured Codex, Claude Code, Cursor, Grok Build, and OpenCode agents while using the user's existing provider subscriptions.

View repository Visit site Mentioned in 31 videos ↓

Overview

The project can be launched with `npx t3@latest`, which starts its backend and local web interface on the machine. It also distributes desktop builds for Windows, macOS, and Arch Linux, and supports remote access from a phone or another machine. The repository describes the project as early-stage and warns that bugs are expected.

What T3 Code is used for

30 uses taken from transcripts — each links to the moment in the video.

  • An open-source control surface for coding agents already running on a developer's computer. It lets users control agents such as Claude Code, Codex, Cursor, Grok Build, and Open Code through mobile, web, or desktop apps while keeping the agents on local hardware.

  • An open-source control surface for AI coding agents running on a user's computer. It lets developers start and steer those agents from a mobile app, web app, or desktop app.

  • Open-source codebase used as the repository for architecture analysis, event-sourcing investigation, game experiments, and an attempted Muse provider integration.

  • Served as the real codebase whose Grok bindings, Cursor integration, and implementation gaps were audited and improved.

  • Provides the repository and coding workflow in which the model audited pull requests.

  • Software project used for pull-request development, remote agent execution, code audits, CI-related work, and autonomous merging.

  • Provides a mobile, web, and desktop control surface for starting, monitoring, and steering multiple coding agents, while running the backend on the user's own machine.

  • T3 Code is a code-related application mentioned alongside VS Code and Cursor, and its mobile app is being rewritten from React Native to SwiftUI. The video discusses working with the app without needing to understand all of its implementation details.

  • Runs remote, multi-threaded AI coding workflows; links threads to pull requests, settles threads when pull requests merge, snoozes work, provides remote development previews, and supports automated interaction with GitHub.

  • Development environment and codebase used for testing AI agents, computer use, code reviews, rewrites, and software changes.

  • Runs code-related agent threads, manages pull requests, supports computer-use workflows, and launches multiple coding tasks in parallel.

  • Coding environment where an agent receives bug screenshots, modifies code, tests changes, creates pull requests, and records videos demonstrating fixes.

  • Agent coding interface used to start repository tasks, inspect thread history and tool calls, run long-lived coding work, and track when a thread needs attention.

  • T3 Code is a software project that uses Clerk for its authentication and billing needs across web, mobile, and desktop platforms. The video also mentions that Clerk built and maintains an Electron package for it.

  • The speaker's own coding agent that calls Claude Code and Codex via the agent SDK to review Claude's work

  • A control surface for AI coding agents already installed on your machine, letting you start, monitor and steer them from mobile, web and desktop apps while the backend runs on your own machine.

  • The speaker's coding agent tool that works with Claude Code and Codex subscriptions, used for coding workflows, PR auditing, and running sub-agents.

    unverified Theo - t3․gg
  • The creator's own coding tool, used throughout as the test environment for running models on audits, PR reviews, delegation, and a new feature that remotely updates machines and providers in one click.

  • An open-source control surface that allows users to steer and drive various coding agents set up on a local machine from a desktop, web, or mobile application.

  • Control surface for driving existing coding agents from desktop, web or mobile apps via a local server

  • The speaker's free, open-source agentic coding tool. Connects to multiple machines from one app or phone, treats the sidebar as an inbox where threads 'settle' when done, and has features like Orchestrator V2, watch pull request, and a working section for hidden threads.

  • Open-source desktop/mobile app for orchestrating AI coding agents across machines, harnesses and subscriptions; the subject of the tutorial

  • Theo's open-source control panel for AI coding agents running on your own machine. Sign into agents like Codex and Claude Code, run one command to start a local server, and manage them from web, desktop, or mobile apps, with permission modes, multiple accounts, and source control.

  • Theo's free, open-source agentic development environment, which he hill-climbed into a complete UI redesign using a multi-model 'council' of reviewer agents. It includes a beta design toggle, an MCP for agent-to-agent coordination, and a sponsor link to Magic Patterns.

  • The host's own coding CLI/agent product. He discusses its token usage telemetry (tens of trillions of tokens per day), its opt-in data-sharing plans, and a FAQ entry describing his multi-account proxy setup.

  • An open-source agent harness control surface that lets developers control coding agents already running on their machine, connecting to Claude Code, Codex, Cursor, Grok Build, OpenCode and Anti-Gravity via local logins. Offers a CLI, web app, and desktop/iOS/Android apps for remote access.

  • The creator's own open-source coding agent tool, used as the case study for handling high PR volume.

  • The creator's open-source coding agent/CLI tool, used throughout the video to run Gemini 4 Carbon via DeepMind's ACP bindings, spin up other models for review, and connect to machines across a fleet.

  • Theo's harness-agnostic coding environment where one harness can call another. It ships fast, includes orchestrator capabilities, and supports features like guest share links for sharing threads with others.

  • An open-source agent harness and control surface from ping.gg for running AI coding agents (Claude Code, Codex, Cursor, Grok, OpenCode, Google Antigravity) on your own machine, with a local server, web app, iOS/Android apps, remote access and permission modes.

Videos mentioning T3 Code

31 in the library.