API Reference¶

Claude Code Proxy provides multiple endpoint modes for different use cases.

Endpoint Modes¶

Claude Code Mode (Default)¶

Base URL: http://localhost:8000/ or http://localhost:8000/cc/
Method: Uses the official claude-code-sdk
Limitations: Cannot use ToolCall, limited model settings control
Advantages: Access to all Claude Code tools and features

API Mode (Direct Proxy)¶

Base URL: http://localhost:8000/api/
Method: Direct reverse proxy to api.anthropic.com
Features: Full API access, all model settings available
Authentication: Injects OAuth headers automatically

OpenAI Codex Mode (Experimental)¶

Base URL: http://localhost:8000/codex/
Method: Direct reverse proxy to chatgpt.com/backend-api/codex
Features: ChatGPT Plus Response API access
Requirements: Active ChatGPT Plus subscription
Documentation: OpenAI Response API

Supported Endpoints¶

Anthropic Format¶

POST /v1/messages         # Claude Code mode (default)
POST /api/v1/messages     # API mode (direct proxy)
POST /cc/v1/messages      # Claude Code mode (explicit)

OpenAI Compatibility Layer¶

POST /v1/chat/completions           # Claude Code mode (default)
POST /api/v1/chat/completions       # API mode (direct proxy)
POST /cc/v1/chat/completions        # Claude Code mode (explicit)
POST /sdk/v1/chat/completions       # Claude SDK mode (explicit)

OpenAI Codex Response API¶

POST /codex/responses                    # Auto-generated session
POST /codex/{session_id}/responses       # Persistent session

Utility Endpoints¶

GET /health              # Health check
GET /v1/models           # List available models
GET /sdk/models          # List models (SDK mode)
GET /api/models          # List models (API mode)

Available Models¶

Claude Models¶

Models available depend on your Claude subscription:

claude-opus-4-20250514 - Claude 4 Opus (most capable)
claude-sonnet-4-20250514 - Claude 4 Sonnet (latest)
claude-3-7-sonnet-20250219 - Claude 3.7 Sonnet
claude-3-5-sonnet-20241022 - Claude 3.5 Sonnet
claude-3-5-sonnet-20240620 - Claude 3.5 Sonnet (legacy)

OpenAI Codex Models¶

Models available with ChatGPT Plus subscription:

gpt-4 - GPT-4 (ChatGPT Plus version)
gpt-4-turbo - GPT-4 Turbo (faster variant)
gpt-3.5-turbo - GPT-3.5 Turbo (base model)

Request Format¶

Anthropic Format¶

{
  "model": "claude-3-5-sonnet-20241022",
  "messages": [{"role": "user", "content": "Hello"}],
  "max_tokens": 1000
}

OpenAI Format¶

{
  "model": "claude-3-5-sonnet-20241022",
  "messages": [{"role": "user", "content": "Hello"}],
  "max_tokens": 1000,
  "temperature": 0.7
}

Codex Response API Format¶

{
  "model": "gpt-4",
  "messages": [
    {"role": "user", "content": "Hello, can you help me?"}
  ],
  "temperature": 0.7,
  "max_tokens": 1000,
  "stream": false
}

Note: The Codex instruction prompt is automatically injected into all requests.

Authentication¶

Claude Endpoints¶

OAuth Users: No API key needed, uses Claude subscription
API Key Users: Include x-api-key header or Authorization: Bearer header

Codex Endpoints¶

ChatGPT Plus Required: Active subscription needed
OAuth2 Authentication: Uses credentials from $HOME/.codex/auth.json
Auto-renewal: Tokens refreshed automatically when expired

Streaming¶

Both modes support streaming responses:

{
  "model": "claude-3-5-sonnet-20241022",
  "messages": [{"role": "user", "content": "Tell me a story"}],
  "stream": true
}

Mode Selection Guide¶

Use Case	Recommended Mode	Endpoint
Need Claude Code tools	Claude Code mode	`/v1/messages`
Need full API control	API mode	`/api/v1/messages`
Using OpenAI SDK with Claude	Either mode	`/v1/chat/completions` or `/api/v1/chat/completions`
Direct API access	API mode	`/api/v1/messages`
ChatGPT Plus access	Codex mode	`/codex/responses`
Session-based conversations	Codex mode	`/codex/{session_id}/responses`