Ralph Workflow
Ralph Workflow is a free, open-source orchestrator for AI coding agents. Hand it a well-specified task, let agents plan, build, verify, and fix, and come back to reviewable, tested work. The full operator manual lives under ralph-workflow/docs/sphinx/.
Who it's for
Ralph Workflow fits developers and small teams with work that is too big to babysit and too risky to trust blindly. It is not for vague prompts or repos without guardrails.
First run
- Install Ralph. Use pipx install ralph-workflow to keep it isolated from
other Python tools. If you do not use pipx, pip install ralph-workflow also works.
- Start in your project. Run cd /path/to/your/project and ralph --init.
It creates your user-global config and a PROMPT.md; project-local config is optional later with ralph --init-local-config.
- Confirm a coding agent. Ralph looks for supported agents already on your
PATH and enables the ones it finds. Install and authenticate an agent first if none are found.
- Check the setup. Run ralph --diagnose and fix any reported problem before
starting work.
- Describe the task. Edit PROMPT.md with the outcome and checks you expect.
For a task-shaped starter, use ralph --init feature-spec, guardrail, refactor, test-coverage, or docs before a prompt file exists.
- Run Ralph. Run ralph, then read the finish-receipt artifact: it names the
change, checks run, and review focus before you decide what to do next.
For the full walkthrough, see Getting started.