pingdotgg/t3code

T3 Code is an agent harness control surface that provides unified control over multiple AI coding agents running on a user's machine.

View on GitHub ↗Jump to charts ↓Open shareable report

Summary Information

Updated 42 minutes ago
Type:ApplicationCategory(s):AI Coding AgentsAI Agents & LLM Apps
Added to GitGenius on August 2nd, 2026
Created on February 8th, 2026
Open Issues & Pull Requests: 2,027 (+1)
GitHub issues: Enabled
Number of forks: 5,952
Total Stargazers: 23,262 (+3)
Total Subscribers: 62 (+0)

Repository Insights (GitGenius)

Median issue/PR response: 19.7 hours
Mean response time: 10.8 days
90th percentile: 35.1 days
Tracked items: 2,641

How this project is maintained

About 16% of issues opened in the past year have never received a reply. 99% of open issues come from outside the core team, so the backlog reflects real-world use rather than internal planning. Work labelled "via-triage" is answered fastest, typically in under an hour, while "🚧 In Progress" waits about 3 days. Almost all tracked open issues have seen activity in the last three months. 75% of issues opened in the past year have been closed, leaving a working backlog.

Charts & Analytics

Fetching additional details & charts...

Issue Activity (beta)

Open issues: 794
New in 7 days: 237
Closed in 7 days: 110
Avg open age: 11 days
Stale 30+ days: 189
Stale 90+ days: 14

Recent activity

Opened in 7 days: 194
Closed in 7 days: 83
Comments in 7 days: 375
Events in 7 days: 1,072

Top labels

  • bug (1,131)
  • needs-triage (974)
  • enhancement (413)
  • via-triage (383)
  • accepted (270)
  • duplicate (114)
  • 🚧 In Progress (42)
  • needs more info (32)

Most active issues this week

Sign in to see which issues are moving.

Detailed Description

T3 Code is an agent harness control surface that provides unified control over multiple AI coding agents running on a user's machine. The project is written in TypeScript and offers access through multiple interfaces: a native mobile app available on iOS and Android, a web application, and an Electron-based desktop application. The core functionality enables users to manage subscriptions and instances of Claude Code, Codex, Cursor, Grok Build, and OpenCode from a single control point, eliminating the need to switch between different agent interfaces.

The project emerged from the creators' desire to build a superior development experience with AI agents. While inspired by existing solutions like the Codex desktop app, Conductor, Claude Desktop, and Cursor Glass, the team felt none of these options met their standards for performance, remote accessibility, and openness.

Installation flexibility is a key design consideration. Users can test T3 Code immediately through a terminal-based server launch without installation, requiring only Node.js 22.16 or later. For persistent use, the desktop application is available through multiple distribution channels including GitHub Releases, Homebrew for macOS, and winget for Windows, with additional support for Arch Linux through the AUR. Before using T3 Code, users must install and authenticate at least one supported provider through their respective CLI tools, whether that's Codex CLI, Claude Code, Cursor CLI, Grok Build CLI, or OpenCode.

The documentation structure reflects the project's maturity level and intended audience. User-focused documentation covers installation, permission modes, keyboard shortcuts, remote access configuration, update management, and source control integrations. Provider-specific documentation addresses multiple account setup for Codex and Claude, while Linux users have dedicated guidance for running T3 Code as a background service. For developers interested in building from source, the documentation begins with an internals overview and requires installation of the Vite+ global command-line tool called vp.

The project explicitly acknowledges its early-stage status, with the creators warning that bugs should be expected. The contribution policy reflects this developmental phase, with the team currently accepting only small fixes while declining larger feature contributions. This conservative approach to external contributions suggests the creators are still establishing the project's direction and architecture. Support for contributors and users is provided through a Discord community server, and detailed contribution guidelines are documented in a CONTRIBUTING.md file that users are directed to read before opening issues or pull requests.