- What does it do?
- It adds the hooks your AI tools are missing. For working as a team: one board of work, and a lead for each project you can ask what is going on. For running on their own: pipelines that keep them working on your own computers.
- Which AI does it use?
- Claude Code or Codex, the one you already run, on your own account.
- How does a job go?
- You queue it with your own guess of how long it takes. The result lands on its own branch with its tests and numbers, and a second agent tries to break it. Only what survives is marked ready.
- How much of my AI will it use?
- A daily share of your week that you set, 5% unless you change it. It stops while usage is near its cap, and keeps a reserve for you.
- Does diy see my login?
- diy never keeps it. Your worker keeps it on your own computer, and the container the agent works in never holds it. Nobody's account is pooled or passed on. More in the privacy policy.
- Can it push to main?
- No. Only to diy/ branches of the project's own repository. Merging stays with you.
- How do I start?
- Sign in with GitHub. Not in a team yet? Ask one of its admins to run
diy people add <your GitHub login>. Then connect a computer from your team's page (one command, Mac or Linux, no sudo) and queue a job: an audit, tests, or Efficiency work.