lidge-jun/opencodex

Universal provider proxy for OpenAI Codex & Claude Code — use any LLM (Claude, Gemini, Grok, DeepSeek, Ollama…) with Codex CLI, App, SDK, and Claude Code

What it solves

OpenCodex is a universal provider proxy that allows users to run any Large Language Model (LLM) within tools that are normally locked to specific providers, such as OpenAI Codex, Claude Code, Claude Desktop, and Grok Build. It removes the need to wait for official support for new models or providers by translating API requests between these tools and a wide variety of LLM backends.

How it works

It operates as a lightweight local proxy that translates the Codex Responses API into the format required by the chosen provider. It supports streaming, tool calls, and reasoning tokens. The proxy can be configured via a web dashboard or CLI, and it can be injected into the Codex CLI, TUI, App, and SDK.

For OpenAI accounts, it manages a pool of ChatGPT accounts to handle authentication and quota management, automatically routing new sessions to the lowest-usage healthy account while maintaining session affinity for existing threads.

Who it’s for

Developers and AI power users who want to use their preferred LLMs (such as Claude, Gemini, Grok, DeepSeek, or local models via Ollama) inside the Codex ecosystem and other AI-driven coding tools without being tied to a single provider.

Highlights

  • Broad Provider Support: Out-of-the-box support for over 40 providers, including Anthropic, Google, xAI, Azure, and any OpenAI-compatible endpoint.
  • Multi-Tool Compatibility: Works with Codex, Claude Code, Claude Desktop, Grok Build, and GitHub Copilot App.
  • Account Pooling: Safely manages multiple ChatGPT accounts with auto-routing based on quota and built-in failover.
  • OAuth Integration: Supports OAuth login for providers like xAI, Anthropic, and Kimi to avoid manual API key management.
  • Subagent Delegation: Allows routing complex tasks to reasoning models and fast tasks to cheaper models via a subagent picker.
  • Enhanced Capabilities: Provides non-OpenAI models with web search and image understanding via a gpt-5.4-mini sidecar.
  • Background Operation: Can be installed as a system service on macOS, Linux, and Windows for seamless background execution.