Ralph Workflow

Install the local autopilot for coding agents, hand it one bounded task, and come back when there is tested work to review.

Ralph Workflow is a free, open-source, local-first loop framework that runs the coding agents you already use unattended on your own machine. Write the spec, run ralph, and decide later whether the result is something you would actually merge.

This page is the maintained product docs home.

Choose your path

• I am brand new → start with Getting Started with Ralph Workflow

• I want the shortest path to a real first run → start with Choose Your First Ralph Workflow Task or Quickstart

• I need configuration or operator detail → start with Configuration Reference or reference

• I want docs grouped by real user goal → start with When Ralph Workflow Fits — and When It Does Not

Docs paths

First run

• Getting Started with Ralph Workflow

• Quickstart

• Choose Your First Ralph Workflow Task

Configuration and customization

• Configuration Reference

• reference

• Advanced Pipeline Configuration

• Advanced Artifact Configuration

• Advanced MCP Configuration

• policy-explanation

Use-case routing

• Which Agent Should I Start With?

• When Ralph Workflow Fits — and When It Does Not

• How to Review AI Coding Output Before You Merge

• What Good Ralph Workflow Output Looks Like

Outside proof surfaces

If you want outside discovery or comparison surfaces before you install, Ralph Workflow also has a few live third-party proof surfaces:

• ToolWise review page — inspect a live third-party review surface without falling back to a repo-first proof path.

• SaaSHub product page — review the product card with the current positioning.

• SaaSHub alternatives — compare Ralph Workflow against adjacent tools in the same browse flow.

Treat those as supporting evidence, not the main path. The best evaluation route is still: inspect the source, choose one real task, run it locally with agents you already use, and decide whether the morning-after result is something you would merge.