OneMux supports four protocol formats: OpenAI, Anthropic, Gemini, and Responses API. Developers can choose the optimal protocol for each model's capabilities and switch without modifying existing code. This page lists the endpoints, authentication methods, and use cases for each protocol.
Supported Protocols
Protocol
Endpoint
Supported Models
Authentication
OpenAI-compatible
/v1/chat/completions
All models
Authorization: Bearer sk-xxx
Native Anthropic
/v1/messages
Claude models
x-api-key: sk-xxx
OpenAI Responses
/v1/responses
GPT models
Authorization: Bearer sk-xxx
Native Gemini
/v1beta/models/{model}:generateContent
Gemini models
Authorization: Bearer sk-xxx
Why Are There Multiple Protocols?
Easy migration: Keep your existing code and change only the Base URL
Full native feature support: Supports features such as Anthropic Prompt Caching, Claude Extended Thinking, and Gemini 2M context
Optimal performance: Claude is 20-30% faster with the Anthropic protocol than with the OpenAI format
Recommended Usage
Model
Recommended Protocol
Reason
Claude models
Native Anthropic
Supports Prompt Caching, reducing costs by 90%
GPT models
OpenAI-compatible
Standard format with the broadest compatibility
Gemini models
Native Gemini
Supports 2M context for processing very long documents