Video summary
T3 Code broke me
Main summary
Key takeaways
Summary
Theo presents T3 Code as an open-source desktop and web app for organizing and running AI coding agents. It began as an alternative to OpenAI’s Codex app after the team became dissatisfied with other options. Theo says it has since developed into a distinct tool rather than simply a weaker copy. At the time of recording, he reported nearly 550,000 users and 140,000 weekly active users.
Main capabilities
- Use multiple AI tools and subscriptions in one place: T3 Code supports a range of coding agents, models, and harnesses, including Codex, Claude, Cursor, Grok, and OpenCode. It also supports ACP-compatible tools and lets users configure multiple accounts or instances of the same provider. Users supply their own subscriptions; T3 Code does not sell AI access.
- Coordinate agents and threads: Agents can launch subagents using different models and harnesses, while threads can communicate with or start work in other threads. T3 Code exposes this orchestration through an MCP server, allowing external tools to manage threads as well.
- Run work across multiple machines: Threads can be assigned to a local computer or remote machine. T3 Code uses WebSockets for its remote architecture, supports connections through Tailscale or T3 Connect, and can route work to machines based on available resources. It also offers remote mobile access, device simulators, and browser sessions.
- Manage work like an inbox: Theo’s central workflow idea is to clear threads when they no longer need attention. Users can settle, snooze, archive, or pin threads; hide active work in a separate section; schedule tasks; and choose whether new messages steer an agent or queue behind its current work. Threads can also link to pull requests and settle automatically when a PR merges.
- Work with code and pull requests: The app includes a PR viewer and monitoring tools, supports several Git providers, and can open projects in VS Code over SSH—even when the project is on another machine. Projects can be transferred between machines, and worktrees are supported for parallel work.
- Monitor and customize: T3 Code provides usage information, machine diagnostics, configurable settings and keybindings, custom themes, and options for handling projects and worktrees. Agents can produce interactive HTML visualizations, and image attachments can be made available to agents as files. A snapshot feature can capture app content and accessibility information for context.
Walkthrough and setup guidance
The video is a hands-on product overview rather than a step-by-step beginner course. Theo demonstrates:
- Installing T3 Code and using existing provider setups.
- Running
npx t3 serveto expose a local instance andnpx t3 connectto link a machine through T3 Connect. - Starting parallel threads on different machines and using multiple agents for a codebase review.
- Using the inbox workflow, scheduled tasks, remote device simulators, PR tools, and the MCP connection.
Theo also describes npx t3 triage, a command intended to give Claude or Codex diagnostic context for troubleshooting local problems and preparing issue reports. The app is available as a stable release, with a nightly channel for users who want newer features and accept more frequent changes.
Assessment and limitations
Theo’s assessment is strongly positive but includes caveats. T3 Code is open source, but users must provide both their own AI subscriptions and the compute it runs on. At the time of recording, T3 Connect was limited to three devices. Built-in computer use was not yet available, though some supported harnesses provide it.
Theo also says parallel workflows can consume usage quickly—not because T3 Code is inherently less efficient, but because it makes it easy to run more work at once. He jokes that the app can encourage late-night “vibe coding.”
Theo says user feedback has been positive, but the video is primarily the founder’s product demonstration and personal review, not an independent evaluation.
Sponsor segment
The video includes a sponsorship for Apify, presented as a source of tools and data for agents. Theo demonstrates using its MCP server with Claude Code through T3 Code to search sources such as YouTube, Reddit, and Hacker News.
Main speakers and sources
- Theo — Host, T3 Code co-founder, and primary demonstrator.
- Julius and Maria — Credited with major contributions to orchestration, interface, and infrastructure work.
- Other T3 Code contributors and users — Referenced for integrations, PR tooling, feedback, and community contributions.
- Apify — Video sponsor.
Rate this summary
Your feedback will help improve summaries.
Improve this summary
Reprocess with a stronger model when the summary feels incomplete or inaccurate.
Translate summary in another language
Ask questions to this video
Chat for follow-up questions, clarifications, and source-backed answers.