What is claudex?

ClauDex is a plugin for Claude Code that pairs Claude with OpenAI's Codex so the two AI coding tools review each other's work instead of just running side by side. When you run the main command, Claude writes code for a task, then Codex reads the change and points out problems, and Claude fixes what it agrees is real. This write and review cycle repeats up to three times, aiming for both models to agree before anything ships. There is also a verdict command that has Claude and Codex look at the same set of changes on their own, without seeing each other's notes, and then compares what each one found. Issues both models flagged are marked one way, issues only Claude caught are marked another way, and issues only Codex caught get a third mark. The result ends in a simple recommendation: ship it, fix it first, or redesign it. A separate debate command lets the two models argue opposite sides of a design choice and produces a short summary of where they landed. The tool only runs when a person types a command, so it never reviews code automatically in the background. It checks that both Claude Code and the Codex CLI are installed and signed in before doing anything, and if Codex is not available it refuses to run rather than pretending a single model review is the same thing. There is also an optional setting that can block commits unless a recent cross review exists, though this is turned off by default in every project until someone turns it on. Setup means installing Claude Code and the Codex CLI, each tied to its own subscription or API key, then adding the ClauDex plugin from its marketplace listing inside a Claude Code session. It is meant to be used inside any project tracked by git, since it reviews the actual changes you have made rather than the whole codebase. This is a good fit for a developer or vibe coder who already uses Claude Code and wants a second opinion from a different AI model before merging changes, without needing to manually copy code between two separate tools.