Skip to main content
Autopilot for your coding agents.
Ralph Workflow
Sign in
Loop Engineering 30+ Projects Open Source

The Ralph Loop Ecosystem

When 30+ independent projects converge on the same pattern — plan, build, verify, recover — it tells you something real about what works for running AI coding agents unattended.

Star Ralph Workflow Add your project

What is the Ralph Loop?

The Ralph Loop is a pattern for running AI coding agents unattended. It has four phases:

  1. Plan — break a spec into bounded, verifiable steps
  2. Build — execute each step in a clean workspace
  3. Verify — check output against the spec before proceeding
  4. Recover — if verification fails, re-plan and retry

Read more in What is Loop Engineering?

Independent Projects Using the Pattern

These projects span multiple languages, agent providers, and use cases — but all converge on the same loop structure. Select projects include:

Ralphex

Multi-agent orchestration framework with planning, dev, review, and fix phases

ralph-addons

Community extensions — custom phases, analyzers, tools, and verification plugins

Ralph-code

Direct loop implementation with plan-build-verify-recover structure and phase routing

Oh-My-OpenClaw

OpenClaw skill framework implementing loop patterns for autonomous AI agent tooling

N43-Studio

Contracts-driven loop execution with `contracts/ralph/core/` defined tasks

Martingale42

Progress-template.json-based loop with structured output and verification

superstack

"AI Superstack: Your CLI to orchestrate agents" — names Ralph as key tool

wtthornton/rloop

Claude Code skill integration with `.claude/skills/ralph-workflow/SKILL.md`

Read the full survey in The Ralph Workflow Ecosystem — 30+ Independent Projects.

Why So Many Projects?

This is convergent evolution — not coordination. When developers independently build tools to solve the same problem (reliable unattended AI coding), they converge on the same architecture because it works. The loop pattern emerges because:

  • Single-shot prompts fail on non-trivial tasks — agents need iteration
  • Unverified output is useless — you need gates between phases
  • Recovery is the hard part — most runners stop at the first failure; a loop retries
  • Vendor lock-in is the hidden cost — a loop abstraction lets you swap the agent underneath

Dive deeper: Why Several Projects Independently Built the Same Loop Pattern

Add Your Project

Is your project using a plan-build-verify-recover loop? Open an issue on the Ralph Workflow repo with your project's name, URL, and a one-sentence description. We'll add it to the ecosystem survey.

Or fork Ralph Workflow and build your own loop — it's free, open-source, and vendor-neutral.

Your loop, your project.

Ralph Workflow is the reference implementation. The loop pattern is yours.

Star on Codeberg Run your first loop