promptadvisers

promptadvisers / claudex

Public

Autonomous Claude + Codex review loop for Claude Code. Plan mode + read-only review mode, all in one window.

33
11
100% credibility
Found Apr 30, 2026 at 33 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Shell
AI Summary

Claudex is a plugin for an AI coding tool that automates iterative debates between two AIs to refine user-submitted feature plans through specialized reviews until approved.

How It Works

1
🔍 Discover Claudex

You stumble upon Claudex, a handy tool where two smart helpers debate your simple project ideas to create rock-solid plans.

2
🛠️ Prepare your setup

You run a quick checker that installs the needed helpers and confirms everything is ready on your computer.

3
📂 Add to your project

You place the tool in your work folder and refresh so it's available right away.

4
Share your idea
🗣️
Chat to refine

Answer a few friendly questions to make your idea crystal clear before starting.

🚀
Dive right in

Launch straight into the action without extra chit-chat.

5
🔥 Watch the magic unfold

One helper drafts a detailed plan while the other challenges it as an engineer, security expert, and operations pro, looping until flawless.

6
📈 Monitor the progress

Check in anytime to see how many debate rounds happened and what issues were fixed.

🎉 Receive your perfect plan

You get a complete, thoroughly vetted plan file that's survived the ultimate test, ready for you to build with confidence.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 33 to 33 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is claudex?

Claudex is a Shell-based plugin for Claude Code, Anthropic's CLI tool, that runs an autonomous Claude agent loop with OpenAI's Codex. You type `/claudex:plan add expiry dates to links`, optionally answer a quick interview, then watch Claude draft a detailed PLAN.md while Codex grills it as a senior engineer, security expert, and ops reviewer—revising until no issues remain, all in one terminal. It solves the problem of weak upfront designs by simulating a multi-persona debate before you code.

Why is it gaining traction?

Unlike solo Claude or basic code reviewers, claudex automates Claude autonomous coding with rotating personas that catch design flaws, security gaps, and ops pitfalls early—delivering a changelog-tracked PLAN.md without manual intervention. Developers dig the fire-and-forget slash commands (`/claudex:plan --rounds 5 `) and utilities like `/claudex:status` or `/claudex:doctor` for zero-babysitting workflows. The read-only review mode adds diff analysis to `reviews/`, making it a lightweight autonomous Claude code companion.

Who should use this?

Backend engineers planning feature migrations or auth overhauls in Claude Code sessions, where you need battle-tested designs fast. SREs and security-focused devs evaluating Claude AI for autonomous work loops before production. Skip if you're not already in the Claude Code terminal flow.

Verdict

Try it if you're deep in Claude Code—solid docs, tests, and fail-open safety make it reliable for daily planning, despite 33 stars and 1.0% credibility signaling early maturity. Maintained version lives in a paid community; v1 on GitHub is a teachable prototype worth forking for custom autonomous Claude agents.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.