One remote control for every coding agent you already pay for

What it does T3 Code runs a server on your machine that fronts the coding agents you’re already signed in to — Claude Code, Codex, Cursor, OpenCode, Grok Build, and a long tail of others, plus anything speaking ACP. Threads sync across desktop, web, and iOS/Android apps, so you can kick off a task at your desk and approve the next command from the bus. It does not resell tokens or proxy your subscriptions; it just drives the CLIs you already authenticated.

The interesting bit It’s deliberately a control plane, not an agent. You can swap models mid-thread and hand a task from Claude to Codex, and each thread gets its own git worktree with diff review, commit, push, and PR in one click. The “what are you selling me?” section answers “nothing” — MIT licensed, fork-friendly by design.

Key highlights

• Works with Claude Code, Codex, Cursor, OpenCode, Grok Build, Antigravity, Kiro, and any ACP agent
• Remote access over local network, Tailscale, or their own T3 Connect
• Native iOS and Android apps, plus desktop and web clients
• Per-thread git worktrees with built-in diff review and one-button PRs
• Performance is treated as a tracked feature — dropped frames count as bugs

Caveats

• Requires at least one agent already installed and signed in on your machine; it’s orchestration, not a standalone tool
• Maintainers state review capacity is limited — features need an approved discussion before PRs

Verdict If you juggle multiple coding agents and want phone access to approval prompts, this is the missing layer. If you use exactly one agent and never leave your desk, it’s extra plumbing you don’t need.