Codex Openai Proxy

Codex OpenAI Proxy

A proxy server that allows CLINE (Claude Code) and other OpenAI-compatible extensions to use ChatGPT Plus tokens from Codex authentication instead of requiring separate OpenAI API keys.

Overview

This proxy bridges the gap between:
- Input: Standard OpenAI Chat Completions API format (what CLINE expects)
- Output: ChatGPT Responses API format (what ChatGPT backend uses)

Features

- ✅ OpenAI API Compatibility: Accepts standard OpenAI Chat Completions requests
- ✅ ChatGPT Plus Integration: Uses your existing ChatGPT Plus tokens
- ✅ Cloudflare Bypass: Handles ChatGPT's Cloudflare protection with browser-like headers
- ✅ HTTPS Support: Works with extensions requiring secure connections (via ngrok)
- ✅ Streaming Responses: Full streaming support for real-time responses
- ✅ CLINE Compatible: Tested extensively with CLINE VS Code extension
- ✅ Array Content Support: Handles both string and array message formats from OpenAI SDK
- ✅ Universal Routing: Bulletproof request routing that bypasses complex warp conflicts

Quick Start

1. Build and Run

2. Setup HTTPS Tunnel (Required for CLINE)

Most VS Code extensions require HTTPS:

Security Note: Always use your own unique ngrok domain. Do not share your domain publicly to prevent unauthorized access to your proxy.

3. Configure CLINE Extension

In VS Code CLINE settings:
- Base URL: https://your-static-domain.ngrok-free.app
- Model: gpt-5 (or gpt-4)
- API Key: Any value (not used, but required by extension)

4. Test Connection

How It Works

Request Flow

1. CLINE → Chat Completions format → Proxy
2. Proxy → Converts to Responses API → ChatGPT Backend
3. ChatGPT Backend → Responses API format → Proxy
4. Proxy → Converts to Chat Completions → CLINE

Format Conversion

Chat Completions Request:

Responses API…