Skip to main content
Documentation
DOCS · CLI REFERENCE

Introduction

Ralph Workflow is a command-line tool that orchestrates AI coding agents through structured, resumable development pipelines.

Introduction

Ralph Workflow is a command-line tool that orchestrates AI coding agents through structured, resumable development pipelines. Instead of ad-hoc AI conversations, Ralph provides a disciplined workflow with planning, development, review, and commit phases.

What Ralph Does

Ralph acts as a pipeline manager that coordinates AI coding agents (like Claude Code, OpenCode, and others) to perform development tasks with:

  • Structured phases - Every task goes through planning, development, review, and commit
  • Automatic checkpointing - Progress is saved so you can resume interrupted work
  • Agent fallback - If one AI agent hits rate limits, Ralph switches to another
  • Built-in verification - Code review phase catches issues before committing

How It Works

The Basic Flow

  1. You describe what you want in PROMPT.md
  2. Ralph generates a plan (PLAN.md)
  3. AI agents execute the plan with automatic verification
  4. Changes are reviewed and committed
# Initialize with a work guide template
$ ralph --init feature-spec

# Edit PROMPT.md with your requirements
$ vim PROMPT.md

# Run the workflow
$ ralph

Key Features

  • Multi-agent support - Works with OpenCode, Claude Code Selector, and more
  • 16+ templates - Pre-built work guides for common tasks
  • Checkpoint & resume - Never lose progress to rate limits or network issues
  • Git integration - Automatic rebase support and conflict resolution

When to Use Ralph

Ralph is ideal for:

  • Larger development tasks that benefit from structured planning
  • Tasks requiring verification where you want AI-powered code review
  • Long-running operations where interruption resilience matters
  • Team workflows where consistent process is important

For quick one-off questions or simple edits, talking directly to an AI agent may be faster. Ralph shines when you need reliability and structure.

Next Steps

Loading…