Introduction

Ralph is a CLI tool that implements the Ralph Methodology — stateless, PRD-driven AI development automated by AI coding agents.

The Problem

Ad-hoc AI prompting produces inconsistent results. Without structure, AI agents wander — exploring too much, skipping tests, making changes that break existing code, or losing context mid-task.

The Solution

Ralph gives your AI agent a repeatable, disciplined workflow:

1. You write a PRD — numbered sections describing what to build
2. You decompose it into tasks — small, dependency-ordered units of work
3. Ralph drives the loop — each iteration boots an agent, assigns a task, enforces red/green TDD, runs quality gates, and commits

Each iteration is stateless: the agent boots from disk, reads the task, implements it with TDD, and commits. No persistent state to corrupt, no context window to overflow.

Why Stateless?

Persistent agents accumulate stale context, drift from instructions, and degrade over long sessions. Ralph’s stateless approach means:

• Every iteration starts clean — the agent reads current state from files, not memory
• Crashes are recoverable — just re-run, the agent picks up where the last commit left off
• Quality is enforced every time — the boot prompt includes TDD rules and quality gates that can’t be forgotten

What You Get

• Automated TDD — failing tests first, then minimum implementation
• Task-driven development — clear progress tracking through task files
• Multi-agent support — works with Claude Code, Gemini CLI, Codex CLI, Continue, and Cursor
• Cost tracking — token usage and dollar costs per task and milestone
• Quality gates — lint, format, typecheck, build, and test must all pass before every commit