Claudecodex
Use Claude Code's UI with OpenAI GPT-5.6 (Sol/Terra/Luna) as the engine — powered by your ChatGPT Plus/Pro subscription via a local Codex proxy. One-line install; plain claude…
Install
git clone https://github.com/karem505/claudecodexInstall command taken from the project's README or registry entry. Always review code before running it.
README excerpt
Read full README on GitHub →claudecodex — Run Claude Code with GPT-5.6 using your ChatGPT subscription
claudecodex lets you use Claude Code with OpenAI models — GPT-5.6 (Sol / Terra / Luna) as the engine — billed to your ChatGPT Plus/Pro subscription through a local Codex proxy, without an Anthropic API key or an OpenAI API key.
You keep the Claude Code agent workflow you like (file edits, terminal flow, approvals, MCP servers, skills) but the model answering is GPT-5.6, running on the ChatGPT subscription you already pay for instead of your Anthropic usage.
One-line install. Plain claude stays completely stock.
How it works
claudecodex launches the real Claude Code UI, points it at a local Anthropic-compatible proxy on 127.0.0.1:18765, and the proxy forwards requests to GPT-5.6 via Codex auth on your ChatGPT subscription:
claudecodex
→ claude (real Claude Code UI)
→ local Anthropic-compatible proxy (127.0.0.1:18765)
→ Codex auth via your ChatGPT Plus/Pro subscription
→ GPT-5.6-sol
→ streamed back into Claude Code
Jump to: Install · Usage & model switching · Configuration · vs alternatives · Troubleshooting · FAQ: Claude Code without an Anthropic API key
Demo
(8× speed — watch the full demo video)
Why run Claude Code with OpenAI models?
- Claude Code UI + GPT models — the best agentic coding interface, driving
OpenAI's strongest coding model (GPT-5.6 Sol at
maxreasoning effort). - Uses the ChatGPT subscription you already have — no pay-per-token API bills; requests go through Codex auth on your Plus/Pro plan.
- Zero interference — a thin launcher that
execsclaudewith process-local environment variables. Your normalclaudecommand, login, settings, and Anthropic models are untouched. - Full 372K context window — configured to GPT-5.6 Sol's real Codex context limit, so the context gauge and auto-compaction are accurate.
- Model switching built in — flip between Sol (best), Terra (mid), and Luna (fast/cheap) per launch or mid-session.
claudecodex is a thin launcher: it starts the proxy if needed, exports the
right environment only for that process, then execs claude.
Requirements
- Claude Code — the
claudeCLI. Install: https://claude.com/claude-code - claude-code-proxy — the local Anthropic⇄Codex proxy by @raine. Needs a build that knows the GPT-5.6 models (v0.1.8+; v0.1.17 or newer recommended).
- A ChatGPT Plus/Pro account, authenticated through the proxy's Codex login.
The proxy keeps its own login (it does not reuse the standalone Codex
CLI's credentials), so you authenticate it once with
claude-code-proxy codex auth login.
This is for personal, local use with your own subscription. Don't expose it as a public API or resell access — routing a ChatGPT subscription through non-official clients is a gray area under OpenAI's terms.
Install
1 — install the launcher (one-liner):
curl -fsSL https://raw.githubusercontent.com/karem505/claudecodex/main/install.sh | bash
2 — install the proxy (skip if you already have v0.1.8+):
Similar plugs
CLI proxy that reduces LLM token consumption by 60-90% on common dev commands. Single Rust binary, zero dependencies
Bash is all you need - A nano claude code–like 「agent harness」, built from 0 to 1
Documented system prompts from Anthropic - Claude Fable 5.1, Opus 5.5, Claude Design, Claude Code. OpenAI - ChatGPT GPT-6-Astra, Codex. Google - Gemini 3.8 Flash, 3.1 Pro,…
Wrap Antigravity, ChatGPT Codex, Claude Code, Grok Build, Muse Code, Devin as an OpenAI/Gemini/Claude/Codex compatible API service, allowing you to enjoy the free Gemini…