Ralph Wiggum CLI by Geoffrey Huntley
Geoffrey Huntley created this framework for AI agent loop back in July 2025 and named it Ralph Wiggum. Back then, it didn't get the traction is deserved, but recently became the talk of the AI-verse when Ryan Carson posted an article about it. And ever since a lot of folks are talking about it and sharing their setups.
Recently, Ian Nuttall built a CLI tool based Ralph Wiggum, and it does seem useful for running agents in loops (works for Claude Code and OpenCode). It's a full-fledged CLI tool and has features like PRD and Skills. You can learn about it more from this discussion on X. In the GitHub repo, the Ralph CLI tool is defined as:
Ralph is a minimal, file‑based agent loop for autonomous coding. Each iteration starts fresh, reads the same on‑disk state, and commits work for one story at a time.
Apart from this, I also found the work of other people about Ralph Wiggum and I will be collecting them here:
- The Ralph Playbook by Clayton Farr
- Ralph for idiots by Agrim Singh
- How to Ralph Wiggum by Geoffrey Huntley
- Ralph for Claude Code by Frank Bria
- Ralph Wiggum, Abundance and Software Engineering by Srinath Krishna
Update:
After publishing this post, I received an email from Ben Williams about the new ralph-tui that he created. And it's an even better tool to orchestrate AI agent loops with coding assistants like Claude Code and OpenCode.
I loved it! And you can learn more about how it works and what it can do from this webpage, and here is some discussion about the same.
That's it.
But I will keep adding more resources and information to this page as I discover something noteworthy.
Webmentions