WindsurfAPI exposes OpenAI, Anthropic, and Gemini compatible APIs backed by a self-hosted proxy that translates requests to Windsurf cloud via gRPC. It supports 100+ models, self-hosted deployment (Docker or npm), and account pooling with rate limiting and failover.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
WindsurfAPI turns Windsurf / Devin Desktop's AI models into three standard API surfaces (OpenAI-compatible, Anthropic-compatible, and Gemini-compatible) via a zero-dependency self-hosted reverse proxy. It exposes HTTP endpoints to translate and forward requests to Windsurf cloud through a local Language Server, manages an account pool, and handles rate limiting and failover.
How it works
- It runs an HTTP service (port 3003) exposing OpenAI and Anthropic APIs.
- Translates incoming requests into Windsurf internal gRPC protocol and communicates with the Windsurf Cloud via a Language Server.
- Maintains an account pool with rate limiting and fault tolerance.
- Strips upstream Windsurf identity before returning results, so models identify as the target API.
Getting started
One-click deployment
git clone https://github.com/dwgx/WindsurfAPI.git
cd WindsurfAPI
bash setup.sh # 建目录 · 给权限 · 生成 .env
node src/index.js
Dashboard: http://你的IP:3003/dashboard
Docker deployment
cp .env.example .env
# (optional) place language_server binary at .docker-data/opt/windsurf/
docker compose up -d --build
docker compose logs -f
Default mounts include data, LS binaries, and workspace directories. See environment variables in the README for customization.
Manual installation
git clone https://github.com/dwgx/WindsurfAPI.git
cd WindsurfAPI
# Language Server binary install
bash install-ls.sh
Recent releases
- v3.9.18 (2026-08-04): v3.9.18
Traction
- Stars: 2926
Behind the repo
- License: MIT
- Language: JavaScript
- Created: 2026-04-09
- Last push: 2026-08-04
- Primary topics include anthropic-compatible, api-gateway, claude, openai-compatible, reverse-proxy, self-hosted
Caveats
- Open issues: 6
- The release notes discuss reasoning and tool-calling behaviors across various models (GLM-5.2, Claude, etc.), with observed fixes and changes in tool usage workflows.
- No explicit license or deployment caveats beyond the MIT license noted in the repository description.