Claude Code Proxy is a proxy server that lets you use Anthropic clients with OpenAI, Gemini, or Anthropic models interchangeably.
The tool solves the problem of client lock-in by intercepting Anthropic API requests and routing them to alternative model providers. It uses LiteLLM as its routing layer and maps Anthropic model names like haiku and sonnet to equivalent models from your chosen provider. The proxy runs as a local server that Anthropic clients connect to, allowing transparent switching between OpenAI, Google Gemini, or direct Anthropic backends without changing client code.
You should adopt this tool if you want to use Claude Code or other Anthropic-compatible clients with models from different providers, or if you need to switch providers without reconfiguring your applications. It suits projects where you have API access to multiple model providers and want flexibility in which backend actually processes requests. The tool supports three configuration modes: defaulting to OpenAI models, using Google Gemini with either static API keys or Application Default Credentials, or passing requests directly to Anthropic. You can customize which specific models map to haiku and sonnet requests, giving you control over cost and capability trade-offs.
The project shows active development with recent commits addressing core functionality. Work has focused on implementing the proxy routing logic and expanding provider support to include Gemini and Vertex AI authentication options. The codebase includes configuration management for multiple authentication methods and model mapping strategies, indicating attention to real-world deployment scenarios.
