Skip to main content
OpenSourceAlternatives

T3 Code

Open source · OSI-approvedStatus: unknownNew

Open-source control plane for coding agents

T3 Code is an open-source control surface for coding agents. T3 Code is an open-source control surface for the coding agents on your machine. It drives your Codex, Claude, Cursor, Grok Build, OpenCode, and Antigravity subscriptions from one desktop, web, and mobile app, and it is free with no paid edition documented.[1][2]

Quick verdict

Free open-source harness driving six coding-agent providers. T3 Code is a free, MIT-licensed control surface that drives Codex, Claude, Cursor, Grok Build, OpenCode, and Antigravity subscriptions from one desktop, web, and mobile app. It runs on your own machine against provider CLIs you install, with model costs billed by those providers.[1][2]

Key features

  • One surface for six agent providers. T3 Code controls Codex, Claude, Cursor, Grok Build, OpenCode, and Google Antigravity agents from a single app. If a provider CLI is installed and authenticated on your computer, T3 Code can drive it.[1][2]

  • Desktop, web, and mobile clients. Clients include an Electron-based desktop app, a local web app, and iOS and Android mobile apps, with remote access from a phone or second machine documented.[1][2]

  • Bring your own subscriptions. There is nothing to buy from T3 Code itself; model usage is billed by the agent providers whose subscriptions you connect.[1][2]

  • MIT-licensed and forkable. The pingdotgg/t3code repository carries an MIT license, and the project states it wants users to hold everything needed to fork if it goes the wrong direction.[1]

  • Wide install channels. Distribution covers a shell install script, a Windows PowerShell installer, an npx trial run, GitHub Releases binaries, winget, scoop, Homebrew, Debian packages, and AUR builds.[1]

Who will love it

  • Developers juggling several agent subscriptions. If you already pay for two or more of Codex, Claude, Cursor, Grok, OpenCode, or Antigravity, one surface beats six terminals. This fit is inferred from the supported-provider list, not stated by the vendor.[1][2]

  • Developers who want agents away from the desk. Remote phone access and mobile apps suit developers who want to check or steer agent runs off-machine. This fit is inferred from the documented remote-access capability.[1][2]

  • Fans of open-source tooling. The MIT license and explicit fork-friendly stance suit developers who require inspectable, forkable developer tools.[1]

Who should avoid it

  • Very early software with expected bugs. The project warns it is very early and to expect bugs, so anyone needing a stable, supported agent platform should wait.[1]

  • No provider account means no agents. T3 Code supplies no models itself; without at least one installed and authenticated provider CLI, there is nothing for it to control.[1]

  • Teams needing SSO, SLAs, or support contracts. No enterprise plan, SLA, SSO documentation, or support contact was found; support points to discussions and Discord. Teams requiring those should look elsewhere.[1][2]

What the reports suggest

Inspected sample: T3 Code GitHub repository README (pingdotgg/t3code), the official t3.codes homepage, and the published OpenAlternative Emdash catalogue record used for the competitor edge. No community review sources were inspected.

1 cited community source. No Reddit originals or G2 reports about T3 Code were inspected during intake research, so no testimonial-grade community quotes are included. The t3.codes homepage shows short vendor-selected X posts; those are marketing quotes, not independent reports, and are not quoted here. Star and fork counts below are October 2026 observations, not live values.

No independent community reports inspected. Intake research inspected the repository README and official homepage only; no Reddit originals or G2 reports were reviewed, so this assessment carries no community experience quotes.[1][2][3]

Vendor homepage quotes are marketing, not evidence. The homepage shows short praise posts from X, but those are vendor-selected social quotes and are not treated as independent reports here.[1][2]

High repository traction at observation time. At observation in October 2026 the repository showed about 27,200 stars and 7,200 forks, which signals attention but says nothing about quality or fit.[1][2]

Where it runs

Desktop platforms covered. Documented install paths cover Windows (winget, scoop), macOS (Homebrew), Debian/Ubuntu (.deb), and Arch Linux (AUR), alongside iOS and Android mobile apps.[1]

Runs against your machine, not a vendor cloud. The t3 command starts a local server and web app with a background-service option; no vendor-operated cloud execution tier was documented.[1]

MCP connectivity documented. Docs cover connecting Claude Code, Codex, ChatGPT, and other agents over MCP, though no stable public API surface was confirmed.[1]

Questions worth asking

How much does T3 Code cost per month?

T3 Code itself is free with no paid edition documented; the README states the team is selling nothing. Model usage is billed by your agent providers, so your monthly cost is whatever your Codex, Claude, Cursor, Grok, OpenCode, or Antigravity subscriptions already charge.[1][2]

Can you self-host T3 Code?

Yes. T3 Code runs on your own machine: the t3 command starts a local server and web app, can run as a background service, and drives the agent CLIs installed locally. Desktop installers cover Windows, macOS, Debian/Ubuntu, and Arch Linux.[1]

Which coding agents does T3 Code work with?

Six providers are documented: Codex, Claude, Cursor, Grok Build, OpenCode, and Google Antigravity. You must install and authenticate at least one provider CLI (for example codex login or claude auth login) before T3 Code can control it; Antigravity needs no CLI and is enabled in Settings.[1]

How is T3 Code different from Emdash?

Both are open-source desktop harnesses that drive installed CLI coding agents instead of separate terminals. Emdash emphasizes parallel agents isolated in Git worktrees with task-queue integrations, while T3 Code emphasizes one control surface across desktop, web, and mobile clients including remote phone access. This comparison is editorial and not endorsed by either vendor.[1][2]

Sources and editorial review

Source research. Reviewed on October 13, 2026. Written by Sanjib Sah. AI assisted; reviewed by a human editor.

We have not run or tested this product.

Inspected sample: T3 Code GitHub repository README (pingdotgg/t3code), the official t3.codes homepage, and the published OpenAlternative Emdash catalogue record used for the competitor edge. No community review sources were inspected. No Reddit originals or G2 reports about T3 Code were inspected during intake research, so no testimonial-grade community quotes are included. The t3.codes homepage shows short vendor-selected X posts; those are marketing quotes, not independent reports, and are not quoted here. Star and fork counts below are October 2026 observations, not live values.

  1. T3 Code GitHub repository README

    official · Observed October 13, 2026

    Repository README for pingdotgg/t3code: agent harness control surface with iOS, Android, web, and Electron desktop apps; supports Codex, Claude, Cursor, Grok Build, OpenCode, and Antigravity; free with bring-your-own subscriptions; install via script, npx, winget, scoop, Homebrew, .deb, and AUR; MIT license; very early project with expected bugs; about 27.2k stars and 7.2k forks observed.

  2. T3 Code official site

    official · Observed October 13, 2026

    Official t3.codes homepage: open-source control plane for coding agents orchestrating Claude Code, Codex, Antigravity, OpenCode, Cursor, and Grok from one surface with bring-your-own subscription; macOS download plus iOS and Android apps; vendor-selected X praise quotes shown.

  3. Emdash catalogue record

    official · Observed October 13, 2026

    Published OpenAlternative Emdash catalogue record: open-source desktop app running multiple coding agents in parallel with automatic CLI detection; used as the competitor edge target.

  4. T3 Code GitHub Discussions

    community · Observed October 13, 2026

    Repository Discussions host Ideas and support threads and the README points to Discord for support. No individual community report was inspected or quoted; listed only to record the community support surface.

Compared with these products

No explicit proprietary baseline is recorded.

Similar projects

Related projects with shared categories, technologies, or licensing.

Orca

OrcaAIOpen

Agent development environment running coding agents in parallel

Stars90.2K
Commitjust now
LicenseMIT

Orca is a free, open-source agent development environment that runs Claude Code, Codex, and other coding agents in parallel, each in its own git worktree.

Pi

PiAIOpen

Extensible terminal AI coding agent with model choice and tree-structured sessions

Stars114.5K
Commitjust now
LicenseMIT

Pi is an extensible terminal AI coding agent with a small core, multi-provider model choice, tree sessions, and a TypeScript extension system.

Jean

JeanOpen

Free open-source desktop workspace for running coding agents

Stars1,321
Commitjust now
LicenseApache-2.0

Jean is a free, open-source desktop app that runs terminal-based AI coding agents from one window, isolating each task in its own git worktree.

OpenCode

OpenCodeAIOpen

AI coding agent for terminal, editor, and desktop

Stars202.9K
Commit8h ago
LicenseMIT

OpenCode is an MIT-licensed AI coding agent available through terminal, desktop, and IDE clients. It supports hosted providers and local models. Code handling depends on the selected provider; optional session sharing uploads conversation history to OpenCode's sharing service.

Kilo

KiloAIOpen

Open-source AI coding agent in the IDE, CLI, and cloud

Stars27.6K
Commitjust now
LicenseMIT

Kilo is an open-source AI coding agent that works inside VS Code, JetBrains IDEs, and Visual Studio, with a command-line build and a cloud runner on the way. It reads the existing codebase and carries out engineering tasks from natural-language instructions, proposing file edits, running commands, and verifying its own work.

SWE-agent

SWE-agentAIOpen

Open source coding agent that fixes GitHub issues with LLMs

Stars20.5K
Commitjust now
LicenseMIT

SWE-agent is an open source coding agent that turns a GitHub issue into a tested fix by letting a language model use shell, editor, and search tools in a sandbox.