What is a Development Autopilot?

A Development Autopilot takes the issues on your GitHub board and returns merged pull requests. Autop runs the coding agent you already use, Claude Code or Codex, on your own machine, runs your checks and an automated review, and merges by rules you set. You decide how often it asks: hands-on, guarded or autonomous.

Who is a Development Autopilot for?

Developers who know what they want built and how it should be checked: solo and indie developers running a product across several repositories, and the team lead who wants a backlog worked without babysitting agent sessions. It is not aimed at people who want an app without touching the stack.

What does it do with an issue?

It takes an issue from your GitHub Projects board, writes a spec, implements it on your machine with your coding agent, runs your checks, opens a pull request, has it reviewed, and merges by your rules. When a task gets stuck it retries a bounded number of times, then hands off to you with a reason.

What is it not?

Not a hosted coding agent: nothing of yours runs in Autop's cloud, and your subscription stays yours. Not a chat-to-app builder: it works from issues in repositories you own, with checks you define. Not a replacement for your coding agent: it runs Claude Code or Codex for you.

What do you need?

• A GitHub organisation you own.
• Claude Code or Codex, installed and logged in.
• A Mac or Linux machine that stays on.

Autop-hosted machines are planned.

How much control do you keep?

All of it, by project: hands-on, guarded or autonomous. Your credentials stay with your coding agent on your machine; the runner never reads, copies or proxies them. Deployment stays with your own pipeline.