lazy documentationv0.94Contents

Start here

lazy runs coding agents on your repository as tasks. Each task works on its own branch, in its own container, and stops for you to review the result, send feedback, and accept it into your branch. Your checkout is never touched while an agent works.

Install

You need git (with user.name and user.email set), Bun and Docker, plus a model credential such as a Claude subscription or an Anthropic API key.

git clone https://github.com/getlazy/lazy.git
cd lazy
bun run install:local

The build checks .bun-version before installing dependencies or generating files. If your default Bun is older and mise is installed, it installs and uses the pinned Bun for this command, leaving your default Bun unchanged. Otherwise it stops immediately, names the running Bun version and path, and prints one command to upgrade Bun and retry. bun run build uses the same check.

This builds lazy and installs it into ~/.lazy/bin; add that directory to your PATH if lazy is not found. Then set it up in a repository that has at least one commit:

cd your-project/
lazy init

Getting started covers these steps in detail, and Credentials explains the model credential.

Where to go next

  1. The dashboard. Open it with lazy dashboard and work from your browser: Dashboard tutorial.
  2. The builder in your terminal. Run lazy builder. The command line's strongest tool is a conversation with a builder that creates, runs and reviews tasks for you: Use a builder, and the builder for how it works.
  3. Lazy Teams (private alpha). The same workflow on a shared team server: Lazy Teams tutorial.

Prefer the command line end to end? Getting started takes one task from creation to merge.