What is gemini-web2api?

gemini-web2api is a Cloudflare Workers adaptation of an existing project that lets you talk to Google Gemini through an API format normally used with OpenAI tools. The idea is to take what Gemini offers through its web interface and make it accessible to software that expects to send requests in the OpenAI style, without paying for a Gemini API subscription. You deploy a single JavaScript file to Cloudflare Workers, which acts as a go-between. When your software sends a request in OpenAI format, the worker translates it into the internal format that Gemini's website uses, sends it to Google, then converts the response back before returning it to your software. The translation works by reverse-engineering the data format that Google's web client uses internally. Several Gemini model variants are supported, including Flash, Flash Thinking (which can produce longer outputs), Pro, Auto, and Lite. The Pro label in this context routes to Flash models unless you have a paid Gemini subscription cookie, so it is a preference label rather than a true model switch. Authentication is optional. If you leave the key list empty, anyone can use your deployed worker without a password. You can add API keys to require a bearer token in the same format OpenAI tools use. The service supports streaming responses, function calling in OpenAI format, and a built-in web search feature that uses Gemini's native search capability. It also exposes a Responses API endpoint for Codex CLI compatibility and a native Gemini API endpoint for the Gemini CLI. Known limitations: image and other multimodal inputs are not supported because uploading files to Gemini requires a proprietary protocol that cannot be replicated through standard HTTP. Each request is also treated as a fresh conversation, multi-turn context is simulated by including prior messages in the prompt rather than maintaining a real session. Google may also rate-limit or block high-volume usage.