I built an MCP server that gives local LLMs access to Google Search, Shopping, Flights, Hotels, Translate, Maps, Weather, Finance, News, Scholar, Books, Images, Trends, and a page fetcher - 14 tools total.

It uses headless Chromium via Playwright instead of the Google Custom Search API, so there's no API key to set up, no usage limits, and you get real Google results (not the CSE subset). Install is just:

pip install noapi-google-search-mcp playwright install chromium

Then point LM Studio or Claude Desktop at it and your LLM can search the web, compare flight prices, check the weather, translate text, look up stocks, etc.

Everything runs locally. The LLM sends a tool call, Playwright opens a headless browser, scrapes the results, and returns structured text.

GitHub: https://github.com/VincentKaufmann/noapi-google-search-mcp PyPI: https://pypi.org/project/noapi-google-search-mcp/

Happy to answer questions about the implementation or MCP in general.