What is gemini-search-mcp?

gemini-search-mcp is a free tool that gives AI assistants like Claude or Cursor the ability to search the web in real time, without needing an API key or paying for a subscription. It works through MCP, a standard protocol that lets AI tools call outside services, and connects to Google's AI Mode search feature, the same Gemini powered search available in the AI Mode tab on Google's website. Instead of calling a paid search API, this project runs a real Chrome browser tab in the background and sends search queries through it directly, using Chrome's developer tools protocol. Because the traffic looks exactly like a normal browser rather than an automated script, Google does not apply the rate limits it normally places on programs making many requests quickly, letting the tool handle over 60 searches per minute for free. The project explains that Google throttles based on how trustworthy a connection's technical fingerprint looks rather than by IP address, which is why a real browser tab avoids the usual blocks. The server exposes two main abilities to a connected AI assistant: a web search tool that returns an answer grounded in current search results, and a more general ask tool that lets the AI Mode system decide on its own whether a web search is needed to answer a question. It also runs an OpenAI compatible chat API on a local port, so tools that expect an OpenAI style interface can use it too. Setup requires Python and a Chrome, Edge, or Chromium browser installed on the machine. If Google shows a CAPTCHA challenge, the README explains how to solve it once in a visible browser window and then reuse that saved browser profile for future headless runs. Known limitations include no memory between separate requests, dependence on Google's page structure for extracting answers, and the general fragility of any tool that relies on browser automation rather than an official API. The project is released under the MIT license.