Universal provider proxy for OpenAI Codex, enabling users to integrate any LLM (Claude, Gemini, Ollama, etc.) with the existing Codex CLI, App, and SDK. It translates Codex API calls to diverse provider formats, allowing users to leverage their preferred LLMs within their familiar Codex workflow, supporting features like streaming and tool calls.

How It Works

This project acts as a lightweight local proxy that intercepts Codex API requests. It uses protocol adapters to translate the Codex Responses API into formats compatible with various LLM providers, including Anthropic Messages, Google Gemini, and OpenAI-compatible endpoints. This approach allows seamless integration of external LLMs without modifying the Codex client, preserving the user's existing workflow.

Quick Start & Requirements

• Install: npm install -g @bitkyc08/opencodex
• Prerequisites: Node.js 18+. Bun runtime is bundled.
• Setup: Run ocx init for interactive configuration.
• Run: ocx start to launch the proxy.
• Docs: Full documentation available at lidge-jun.github.io/opencodex.

Highlighted Details

• Broad Provider Support: Integrates 40+ LLM providers via 5 protocol adapters, including custom OpenAI-compatible endpoints.
• ChatGPT Account Pooling: Manage multiple accounts, refresh quotas, and auto-route new sessions to lower-usage accounts while maintaining thread continuity.
• Flexible Authentication: Supports OAuth (Anthropic, xAI), API keys, and environment variables.
• Seamless Codex Integration: Injects into Codex CLI, App, SDK; routed models appear natively in the Codex App model picker.
• Enhanced Capabilities: Non-OpenAI models gain web search/image understanding via a gpt-5.4-mini sidecar.
• Web Dashboard: GUI for managing providers, auth, models, and monitoring requests.
• Background Service: Installable as a system service (launchd, systemd, Task Scheduler) for persistent operation.

Maintenance & Community

The project is community-maintained, with no specific links to community channels, roadmaps, or notable contributors provided in the README.

Licensing & Compatibility

• License: MIT.
• Compatibility: Permissive MIT license allows commercial use. However, users are cautioned that some providers may restrict API traffic routed through third-party proxies, requiring review of provider Terms of Service.

Limitations & Caveats

Cursor integration is experimental. Users must be aware that certain LLM providers may suspend accounts for using third-party proxies. Remote access requires explicit configuration and authentication.