Ralph Workflow is a free, open-source AI agent orchestrator for coding work. You give it one well-specified task; it runs a Ralph loop (plan → build → verify → fix) with your chosen coding agent, then you come back to inspect the result. That simple core ships with a strong default workflow for writing software — adopt it as-is first, then extend it when you need to.
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 tests or other guardrails.
Next step
For the shortest honest first run — install, diagnose, one focused task,
and what success looks like — open START_HERE.md.
The fuller tutorial is
Getting started.
The full operator manual lives under
ralph-workflow/docs/sphinx/.
Supported agents
Ralph Workflow ships eight built-in agent backends: Claude Code,
Claude Code headless, Codex, OpenCode, Nanocoder, AGY (Google Anti
Gravity), Pi, and Cursor. Pick one, authenticate it once on your
machine, and Ralph Workflow uses it. Selection and trust-boundary
details are in agents and
agent-compatibility.
Documentation route
START_HERE.md— runnable first-run walkthroughdocs/README.md— docs map by reader questionralph-workflow/docs/sphinx/index.rst— maintained operator manual (configure, operate, extend)
Project home
- Runtime: Python ≥ 3.12, local-first.
- License: AGPL-3.0-or-later.
- PyPI: https://pypi.org/project/ralph-workflow/
- Issue tracker: https://github.com/Ralph-Workflow/Ralph-Workflow/issues/new
- Contribution route:
CONTRIBUTING.md→ralph-workflow/CONTRIBUTING.md
The Ralph Loop pattern is attributed to Geoffrey Huntley (ghuntley.com/ralph); Ralph Workflow is an independent reference implementation.