freellmapi-proxy — Claude Code skill: one proxy for free LLMs

What freellmapi-proxy does

FreeLLMAPI gathers keys from roughly fourteen providers — Google, Groq, Cerebras, SambaNova, NVIDIA, Mistral, OpenRouter, GitHub Models, Hugging Face, Cohere, Cloudflare, Zhipu, Moonshot, MiniMax — behind a single /v1/chat/completions. Send model: "auto" and the router picks whichever provider is available; on 429 or 5xx it moves down the fallback chain. Provider keys sit in SQLite encrypted, and multi-turn conversations can stick to one provider.

Setup is a plain Node project: git clone, npm install, then an ENCRYPTION_KEY in .env. That key is a 64-character hex string and it protects every stored provider key, so the store is useless without it. npm run dev starts the server plus a Vite dashboard; the production build serves the API and panel on port 3001.

You add keys on the Keys page, reorder providers under Fallback Chain, and copy the unified freellmapi-… bearer token from the same page. Any OpenAI SDK client works — point base_url at http://localhost:3001/v1. Tool calling runs across all providers, and Gemini requests get translated to functionDeclarations and back on the fly.