Skip to content
Scarn
scarn.io

Introduction

Scarn runs AI agents over your backlog, on your machines. You say what should come out the other end, what evidence you need to trust it, and what must never be touched; the agents work out the path. A worker on your laptop or CI box runs a pipeline against each work item that arrives.

The first pipeline is a default bug pipeline. For each bug report it checks there’s enough to act on, then investigates and reproduces the bug, writes a failing test, and proposes a fix. What you get back is an investigation report, a Playwright trace when the app is reachable, the test, and the fix. The designer adapts it to your repository, using the MCP servers and skills it finds there.

A work item arrives, a pipeline is picked for it, a worker runs the current step, and the pipeline’s transitions decide whether to advance, wait for a human, or finish.

You’ll need

Section titled “You’ll need”
  • A git repository with a remote — see Installation.
  • Node.js or Bun — see Installation.
  • Claude Code, Codex, or GitHub Copilot, or an API key from any AI provider — see Installation.

Where to go next

Section titled “Where to go next”
QuickstartTen minutes to a running pipeline.
Core conceptsThe dozen words the rest of the docs assume.
CLI referenceEvery scarn command and flag.

Start now

Section titled “Start now”
Terminal window
npm i -g @scarn/cli && scarn init