macOS + Linux

AI agents clear your backlog while you sleep.

Load cards onto a Kanban board. 3xo spins up Claude Code, Codex, or Gemini, patches your repo, and posts every diff.

Beta is open. Watch the demo →

If the player is blocked, watch on YouTube.

Powered by AI

Claude Code, Codex, or Gemini CLI.

Choose which coding assistant handles each task. Switch between agents or run them in parallel across different cards.

Private by default

All work stays on your machine.

Code, credentials, and session logs live in local SQLite. No cloud sync, no third-party servers.

Flexible workflow

You pick the review model.

Agents write and review each other, you write and agents review, or fully automated. Configure per card or globally.

How it works

Stack the board, let agents run.

  1. Connect your repo. Grant 3xo access once and share commands, secrets, and tests inside each card.
  2. Stack the board. Drag cards into Ready with context and exit criteria so agents know what "done" means.
  3. Let runs finish. The coding agent ships the patch, the reviewer checks it, and you merge.
Watch demo →

Built for people who ship nightly.

Solo devs and lean teams stay unblocked without babysitting every commit.

  • Claude Code CLI
  • Codex CLI
  • Gemini CLI
  • Light and Dark modes
  • Configurable concurrency
  • Complete audit trail

Parallel by design

Each card runs in its own git worktree. Agents can build, test, and patch multiple cards at once without clobbering your main checkout.

Frequently asked questions

Quick answers about the beta.

3xo works with Claude Code CLI, Codex CLI, and Gemini CLI. Assign agents by task type: backend, frontend, planning, docs. If one fails, 3xo hands the card to your backup automatically.

3xo runs on macOS and Linux. It's a native desktop app with a modern UI, and it keeps all data local in SQLite on your machine.

A paired agent reviews every completed card. P0/P1 findings send the work back with notes; clean runs move to Done. You control who reviews what or turn review off.

Yes. Set global and per-project concurrency caps so several agents can hustle without cooking your laptop. Start with one agent per repo, then scale.

Each run keeps a session log, so agents remember previous steps and multi-part tasks stay on track. Every action lands in the board history.

3xo is free during the beta. Install it, run it on real projects, and send feedback. We'll publish pricing before launch.

It's pronounced "exo" like "exo suit." The 3 is a l33t "e".

Release History

Track our progress as we build in public.

View all releases →

Loading changelog...