dwgx/WindsurfAPI

What it solves

WindsurfAPI provides a way to access over 100 AI models (including Claude, GPT, Gemini, DeepSeek, and others) by converting the internal protocols of Windsurf (formerly Codeium/Devin Desktop) into standard, compatible APIs. This allows users to use these models within their own SDKs or AI agents like Claude Code, Cline, and Cursor without needing to use the official Windsurf IDE.

How it works

The project acts as a protocol translation layer. It exposes HTTP endpoints that are compatible with OpenAI, Anthropic, and Gemini API formats. When a request is received, WindsurfAPI translates it into the internal gRPC protocol used by Windsurf, sends it through a local Language Server (LS) binary to the Windsurf cloud, and then translates the response back into the requested standard format. It also manages an account pool to handle load balancing, rate limiting, and failover.

Who it’s for

• Developers who want to use high-end AI models via standard APIs in their own applications.
• AI Agent users (e.g., Claude Code, Cline, Cursor) who want to integrate Windsurf's model access into their existing workflows.
• Self-hosters looking for a unified gateway to multiple frontier LLMs.

Highlights

• Multi-API Compatibility: Simultaneously supports OpenAI, Anthropic, and Gemini API standards.
• Extensive Model Support: Access to 100+ models including various versions of Claude, GPT, Gemini, and specialized models like SWE.
• Account Management: Built-in account pooling with automatic polling, rate limit isolation, and fault tolerance.
• Zero Runtime Dependencies: Built with pure Node.js and requires no npm runtime dependencies.
• Session Management: Includes a server-side session store for the Responses API to maintain conversation state without re-sending full history.