API Overview

Anthropic-compatible API for Claude and Gemini models

Base URL

http://localhost:8080

The default port is 8080. You can configure this via the PORT environment variable or in config.json.

API Compatibility

Antigravity Claude Proxy provides an Anthropic Messages API-compatible interface. This means you can use the proxy as a drop-in replacement for the official Anthropic API by:

1. Setting ANTHROPIC_BASE_URL=http://localhost:8080
2. Setting ANTHROPIC_AUTH_TOKEN=test (or any value if API key auth is disabled)

The proxy translates requests from Anthropic format → Google Generative AI format → Antigravity Cloud Code API, then converts responses back to Anthropic format.

Request/Response Format

All API requests use JSON payloads with Content-Type: application/json.

Responses follow the Anthropic Messages API format:

{
  "id": "msg_01...",
  "type": "message",
  "role": "assistant",
  "content": [...],
  "model": "claude-sonnet-4-5-thinking",
  "stop_reason": "end_turn",
  "usage": {
    "input_tokens": 150,
    "output_tokens": 250
  }
}

Available Endpoints

Messages

Send messages to Claude or Gemini models

Models

List available models

Health

Check server and account health

Authentication

Configure API authentication

Streaming Support

The proxy supports Server-Sent Events (SSE) for streaming responses. Set stream: true in your request to receive incremental updates as the model generates its response.

See the Messages endpoint for details.

Model Support

Claude Models

• claude-opus-4-6-thinking - Extended reasoning, highest capability
• claude-sonnet-4-5-thinking - Balanced reasoning and speed
• claude-sonnet-4-5 - Fast, production-ready

Gemini Models

• gemini-3.1-pro-high - Highest quality Gemini
• gemini-3.1-pro-low - Cost-optimized Gemini
• gemini-3-flash - Fastest Gemini model

Use GET /v1/models to see all available models in your accounts.

Error Handling

Errors follow the Anthropic API format:

{
  "type": "error",
  "error": {
    "type": "invalid_request_error",
    "message": "messages is required and must be an array"
  }
}

Error Types

• authentication_error - Invalid or missing API key (if auth is enabled)
• invalid_request_error - Malformed request or invalid parameters
• api_error - Upstream API error or network issue
• not_found_error - Endpoint not found

Rate Limiting

The proxy automatically handles rate limits across multiple accounts:

• When an account is rate-limited, the proxy switches to another available account
• Rate limit state is tracked per model
• Accounts are automatically re-enabled when their quota resets

You can monitor account status via the Health endpoint or the Web UI.