All skills
Skillintermediate

Quick Start

Run your first Ralph orchestration in about 10 minutes.

Claude Code Knowledge Pack7/10/2026

Overview

Quick Start

Run your first Ralph orchestration in about 10 minutes.

1. Install Ralph

If you haven't installed Ralph yet, follow the full Installation guide.

Quick install (npm):

npm install -g @ralph-orchestrator/ralph-cli

2. Install a Backend CLI (Claude Recommended)

Ralph needs at least one AI CLI tool available on your PATH.

# Claude Code
npm install -g @anthropic-ai/claude-code

# Verify the CLI is available
claude --version

If the backend requires authentication, complete its login flow per the provider's instructions.

3. Verify Setup with ralph doctor

Run the doctor command to validate your environment:

ralph doctor

Fix any WARN or FAIL items before continuing. If you see auth warnings, verify your backend CLI is logged in.

4. Initialize a Project

mkdir my-ralph-project
cd my-ralph-project
git init  # Ralph works best with git

# Create a default config
ralph init --backend claude

This creates ralph.yml in your project.

5. Create a Minimal Hat Collection

Ralph can run with hats (role-based personas) for more structured workflows. Create a minimal hat collection file:

# hats.yml
event_loop:
  starting_event: "task.start"

hats:
  builder:
    name: "Builder"
    triggers: ["task.start"]
    publishes: ["task.done"]
    instructions: |
      Implement the task from PROMPT.md.
      Run any relevant tests.
      When finished, emit task.done and print LOOP_COMPLETE.

6. Define Your Task

Create a PROMPT.md file with your task:

# Task: Create a Todo List CLI (Rust)

Build a Rust command-line todo list with:
- Add tasks
- List tasks
- Mark tasks complete
- Save to a JSON file

Include error handling and unit tests.

7. Run Ralph

# Traditional mode (uses ralph.yml)
ralph run

# Hat-based mode (uses hats.yml)
ralph run --config hats.yml

# Inline prompt example
ralph run -p "Add input validation to the user API endpoints"

8. Understand the Output

While running, Ralph shows a TUI with:

  • Current iteration number
  • Elapsed time
  • Active hat (if hat-based)
  • Recent agent output

Ralph stops when one of these occurs:

  • LOOP_COMPLETE is output (success)
  • Maximum iterations reached (default: 100)
  • Maximum runtime exceeded (default: 4 hours)
  • You quit the TUI

When it finishes, review the generated files in your project directory and .agent/ run logs.

Command-Line Options

# Limit iterations
ralph run --max-iterations 50

# Use different config file
ralph run -c custom-ralph.yml

# Resume interrupted session
ralph run --continue

# Quiet mode for CI
ralph run -q

Example Tasks

Simple Function

Write a TypeScript function that validates email addresses.
Include unit tests.

Web Scraper

Create a web scraper that:
1. Fetches the Hacker News homepage
2. Extracts the top 10 stories
3. Saves them to JSON

Use Node.js with a simple HTML parser.

CLI Tool

Build a markdown to HTML converter:
- Accept input/output file arguments
- Support basic markdown syntax
- Add --watch mode

Next Steps