Repolode
Submit
lidge-jun

lidge-jun/opencodex

Local proxy that lets OpenAI Codex and Claude Code run on any LLM provider, including Claude, Gemini, Grok, DeepSeek, and Ollama.

View on GitHubWebsite
Stars
17,565
Forks
1,351
Open issues
103
Last commit
yesterday

What is it?

opencodex is a local proxy that translates the OpenAI Responses API used by Codex into the protocol of whatever provider you choose. Streaming, tool calls, reasoning tokens and images are translated in both directions, so Codex CLI, the Codex app, Claude Code, Claude Desktop and Grok Build keep their native interfaces while the model behind them can be Claude, Gemini, Grok, GLM, DeepSeek, Kimi, Qwen, Ollama or another OpenAI-compatible endpoint.

Setup is done through a web dashboard on localhost:10100 or the ocx CLI. It ships with 40+ built-in providers and supports routing a model with a provider/model syntax, combos that fail over or round-robin across providers, and sub-agents that run on routed models. It can also pool multiple ChatGPT accounts for Codex auth, with quota-aware routing and session affinity.

The CLI runs on Node 18+ and is installed with npm. Desktop builds exist for macOS, Windows and Linux, and a Docker Compose setup is provided for remote hosting with a bearer token.

Key features

  • 01Translates Codex Responses API calls to Claude, Gemini, Grok, DeepSeek, Ollama and other providers.
  • 02Keeps streaming, tool calls, reasoning tokens and images working in both directions.
  • 03Provides a web dashboard for managing providers, models and accounts.
  • 04Supports combos for failover or weighted round-robin across providers.
  • 05Pools ChatGPT accounts with quota-aware routing and session affinity.
  • 06Offers CLI, desktop app and Docker Compose installation.

Use cases

  • Running Codex CLI or the Codex app on a non-OpenAI model such as Claude or Gemini.
  • Using Claude Code with a cheaper or local model served by Ollama or DeepSeek.
  • Failing over between several providers when one hits rate limits.
  • Hosting a shared proxy on a server with a bearer token for a small team.

Topics

#llm-proxy#codex#claude-code#openai-compatible#ollama#multi-model#developer-tools#ai-gateway#ai-tools#anthropic

FAQ

What is opencodex?
Local proxy that lets OpenAI Codex and Claude Code run on any LLM provider, including Claude, Gemini, Grok, DeepSeek, and Ollama. opencodex is a local proxy that translates the OpenAI Responses API used by Codex into the protocol of whatever provider you choose. Streaming, tool calls, reasoning tokens and images are translated in both directions, so Codex CLI, the Codex app, Claude Code, Claude Desktop and Grok Build keep their native interfaces while the model behind them can be Claude, Gemini, Grok, GLM, DeepSeek, Kimi, Qwen, Ollama or another OpenAI-compatible endpoint.
Is opencodex free to use? What license does it have?
opencodex is open source under the MIT license. Check the license terms before using it in commercial products.
Is opencodex still maintained?
Actively maintained. The last commit to opencodex was on Oct 18, 2026.
What can I use opencodex for?
Running Codex CLI or the Codex app on a non-OpenAI model such as Claude or Gemini. Using Claude Code with a cheaper or local model served by Ollama or DeepSeek. Failing over between several providers when one hits rate limits. Hosting a shared proxy on a server with a bearer token for a small team.
Where can I get opencodex?
The source code and installation instructions are on GitHub: https://github.com/lidge-jun/opencodex

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects