HarimxChoi/google-surf-mcp

google-surf-mcp combines Google search, page fetching, and academic-paper extraction behind five MCP tools. The default abstract mode is a smart way to limit token usage before requesting full documents, while support for HTML and PDFs from sources such as arXiv, OpenReview, and PubMed makes the server useful for research workflows. Local browser profiles, CAPTCHA handoff to a human, cache controls, an SSRF guard, disabled telemetry by default, and a fail-fast cloud mode show that the project has considered real operating conditions rather than only the happy path.

The main risk is dependence on browser automation against a frequently changing search interface. Parser accuracy, regional differences, rate limits, sponsored-result filtering, and Google’s terms can all affect reliability. The project should publish a reproducible regression corpus, clearer accuracy metrics, and a maintenance policy for automated repair PRs. Users also need precise guidance on profile data, cookies, cache retention, and the risks of stealth techniques. This MCP is best for local research assistants that need unified search and extraction without an API key, provided operators accept occasional CAPTCHA intervention and monitor upstream compatibility.