Now on Mac · free, no account
Your Claude Code,
now a whole crew.
Open your repo and describe a goal. A crew of agents plans it, builds it in parallel, tests and reviews it, then hands you one change to approve.
Apple Silicon build · Download for Intel
Free · macOS on Apple Silicon or Intel · needs Claude Code or Codex, signed in
Add a DELETE /notes/:id endpoint with a test. Match how the other routes are done.
Captain
A backend specialist is on it: it'll follow the existing route and test conventions, then a tester and the CodeReviewer check it before it comes to you.
feat: add DELETE /notes/:id endpoint with test
204 on success, 404 when the note doesn't exist, following the existing route and test conventions. v2 fixes the reviewer's finding: a client-supplied id could override the generated one.
Approve commits it to your repo. Request changes sends it back with your note. Nothing ships until you say so.
How it works
A goal in. A reviewed change out.
- 1
Open a folder
Point OpenCrew at a repo, the way you would in your editor. No setup, no sign-in: it uses the Claude Code or Codex you are already logged into.
File → Open Folder… ⌘O
- 2
Give it a goal
Say what you want done, in a sentence. The Captain plans it and puts specialists on the pieces, each in its own git worktree, all at the same time.
Captain → backend-1 · tester-1
- 3
Approve the result
You get one card: the files changed, tests passing, and the reviewer’s verdict. Approve & commit, or send it back with a note.
✓ Tests 5/5 · ✓ Reviewed
Why OpenCrew
A crew, not a chat.
A chat tool gives you one agent and a transcript to babysit. OpenCrew gives you a team, and a result to approve.
Runs on the subscriptions you already have
Agents use your Claude Code and Codex logins. No API keys, no OpenCrew account, no new bill. Pick the harness per agent.
- CaptainClaude Code
- backend-1Codex
- tester-1Claude Code
- CodeReviewerClaude Code
Many agents, in parallel
Specialists work at the same time, each in its own isolated git worktree, so they never trip over each other or over your working copy.
Reviewed before you see it
Every change is tested and read by a CodeReviewer agent first. Send-backs happen between agents, not in your inbox.
Nothing ships without you
Your repo is untouched until you press Approve & commit. Request changes and it goes back with your note.
Local and private
It runs on your Mac and keeps its state in your folder. Your code goes only to Anthropic or OpenAI, through your own login.
FAQ
Questions, answered
Is it free?
Yes. OpenCrew is free to download and use. The agents run on your own Claude Code or Codex subscription, so the work counts against the plan you already pay for.
Do I need an account?
No. No sign-up, no sign-in, no API keys. Install it, open a folder, describe a goal.
What does it send, and to whom?
Your code and prompts go only to Anthropic or OpenAI, through your own Claude Code or Codex login, exactly as when you use those tools directly. OpenCrew itself receives anonymous usage stats (which features ran and whether they worked), never code, prompts, file paths or names, and you can switch that off in Settings. The details are in the privacy policy.
Which models and harnesses does it use?
Claude Code and Codex, the agent tools you already have installed. You choose the harness per agent, and each one uses the models your own setup gives it.
What happens to my repo?
Agents work in isolated git worktrees outside your folder, so your working copy and branch change only when you approve. OpenCrew keeps its record in a .opencrew/ folder in your repo, with its local state git-ignored.
Does it run on Intel Macs? Windows? Linux?
Intel and Apple Silicon Macs both have native builds; the download button picks yours, or grab the Intel build directly. Windows and Linux aren't supported yet.
Put a crew on your next goal.
Download it, open a repo, and get your first reviewed change in minutes.
Apple Silicon build · Download for Intel
Free · macOS on Apple Silicon or Intel · needs Claude Code or Codex, signed in