Gemini Web2API - Free OpenAI-compatible API for Gemini — no keys, no limits.
by•
Turn your free Gemini web account into a full OpenAI-compatible API — local, free, no API keys. One Python file serves /v1/chat/completions, /v1/responses and /v1beta with SSE streaming, tool calling and image support, so any agentic client (AionUI, Claude Code, Cursor, LangChain, Dify) works instantly. Self-healing: auto-updates Google's build label, falls back through proxies on 429, auto-refreshes cookies via a companion extension, and recovers the XSRF token on fresh boot. MIT, zero cost.
Replies
I built this out of frustration: I wanted to ship on Gemini but couldn't justify the API cost, and every free bridge I tried died within days, Google rotates its build label (instant 405s), rate-limits bursts (429s), and stale cookies silently return garbage. So Gemini Web2API is built to be boringly reliable: it auto-updates the build label, falls back through your proxy chain, auto-refreshes cookies from a tiny companion extension, and even recovers the XSRF token on a fresh boot, no config, no babysitting. One Python file, start_server.bat, point your client at http://127.0.0.1:8081/v1 and you have streaming, tool calling and image support for free. It's MIT-licensed and 100% local, your prompts never leave your network. If it saves you money or you build something with it, drop a star and tell me what you'd add next.