Use GPT, Gemini, Grok, DeepSeek, Ollama, or any other LLM with Claude Code — without waiting for Anthropic to add support.

openclaudecode is a lightweight local proxy that translates Claude Code's Anthropic Messages API into whatever your provider speaks. Streaming, tool calls, reasoning tokens, images — everything works, in both directions.

Claude Code CLI ──/v1/messages──▶ openclaudecode ──▶ Any provider
                                        │
              OpenAI · Google · xAI · DeepSeek · Ollama · Groq
              OpenRouter · Together · Fireworks · Mistral · …

Quick start

# Install
npm install -g @bitkyc08/openclaudecode

# Interactive setup
occ init

# Start the proxy
occ start

# Use Claude Code through the proxy
ANTHROPIC_BASE_URL=http://127.0.0.1:10200 claude

How it works

Claude Code sends requests in Anthropic's Messages API format. openclaudecode intercepts them and translates to the target provider's native format:

• OpenAI/xAI/DeepSeek/Groq/Ollama → Chat Completions API
• Google Gemini → Gemini generateContent API
• Anthropic (other key/model) → Messages API passthrough

Tool calls, streaming, images, and thinking/reasoning tokens are all translated bidirectionally.

Configuration

Config lives at ~/.openclaudecode/config.json:

{
  "port": 10200,
  "defaultProvider": "openai",
  "providers": {
    "openai": {
      "adapter": "openai-chat",
      "baseUrl": "https://api.openai.com/v1",
      "apiKey": "${OPENAI_API_KEY}",
      "defaultModel": "gpt-5.5",
      "models": ["gpt-5.4", "gpt-5.4-mini"]
    },
    "google": {
      "adapter": "google",
      "baseUrl": "https://generativelanguage.googleapis.com/v1beta",
      "apiKey": "${GEMINI_API_KEY}",
      "defaultModel": "gemini-3.1-pro"
    },
    "ollama": {
      "adapter": "openai-chat",
      "baseUrl": "http://localhost:11434/v1",
      "defaultModel": "llama3.3"
    }
  }
}

Model naming

Use <provider>/<model> to route explicitly, or just the model name for auto-routing:

# Explicit provider
ANTHROPIC_BASE_URL=http://127.0.0.1:10200 claude --model openai/gpt-5.5

# Auto-routed (gpt- prefix → openai provider)
ANTHROPIC_BASE_URL=http://127.0.0.1:10200 claude --model gpt-5.5

# Local model
ANTHROPIC_BASE_URL=http://127.0.0.1:10200 claude --model ollama/llama3.3

CLI

occ init                       # interactive setup
occ start [--port 10200]       # start the proxy
occ stop                       # stop the proxy
occ status                     # check if running
occ help                       # show help

Roadmap

• [ ] Web dashboard (port from opencodex)
• [ ] OAuth support (xAI, Kimi)
• [ ] Web search sidecar
• [ ] Vision sidecar
• [ ] Background service (launchd/systemd)
• [ ] ocx start --claude unified mode with opencodex

Related

• opencodex — same idea, but for OpenAI Codex CLI

License

MIT