Try Codex's gpt-5.6-sol model inside the Claude Code interface you already use daily.
Run a local, checksum verified proxy that keeps your model traffic on your own machine.
Switch between Claude Code and a Codex backed session with a single command.
Diagnose or repair a broken proxy setup using the built in doctor and setup commands.
| liuzhao1225/claudex | ai-engineer-skool/zen-agentic-engineer-config | ali-expandings/mactune | |
|---|---|---|---|
| Stars | 16 | 16 | 16 |
| Language | Shell | Shell | Shell |
| Setup difficulty | moderate | easy | easy |
| Complexity | 3/5 | 2/5 | 1/5 |
| Audience | developer | developer | general |
Figures from each repo's GitHub metadata at analysis time.
Unofficial workaround that Anthropic does not support, the README warns it may put your account at risk, and it requires a ChatGPT account with Codex access.
Claudex lets you run OpenAI's Codex subscription models, particularly one called gpt-5.6-sol, inside the Claude Code terminal interface. It works by starting a small private proxy on your own machine that Claude Code talks to instead of Anthropic's servers, so from your point of view you keep using the familiar Claude Code interface while the model answering you is actually a Codex model reached through your ChatGPT account. The proxy only listens on localhost, meaning nothing outside your own computer can reach it. Installation is a single command that downloads a checksum verified installer, sets up a private configuration folder, opens a browser window for you to log into Codex through OAuth, and confirms the target model is reachable before finishing. After that you simply run the claudex command instead of claude to launch Claude Code through the proxy. A collection of subcommands handles ongoing use, including claudex setup to repair a broken install, claudex login to redo authentication, claudex doctor for diagnostics, claudex start, stop, and status to manage the background proxy process, and claudex uninstall to remove everything. The project is careful about what it touches on your system. It keeps its files under standard user-scoped directories, does not modify your existing Claude Code or Codex configuration, does not read your saved Codex login credentials, and will add at most one small reversible line to your shell configuration if needed, with a backup made first. The proxy's local API token is stored in a file only you can read and is never printed to the screen. The README is unusually direct about the risk involved: this is an unofficial, community built workaround, and Anthropic does not support routing Claude Code to non-Claude models this way. Using it could put your account at risk, and the project explicitly points to that warning before telling you to install it. It runs on macOS or Linux, requires Claude Code already installed, and needs a ChatGPT account with Codex access. It is released under the MIT license.
A one-command tool that lets you run OpenAI's Codex models inside the Claude Code interface through a private, localhost-only proxy on your own machine.
Mainly Shell. The stack also includes Shell, Bash, CLIProxyAPI.
Free to use, modify, and redistribute for any purpose, including commercial use, as long as you keep the copyright notice.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.