Hey Product Hunt! We kept running into the same thing: plain requests to Google mostly get /sorry/ captchas now, and SERP APIs get expensive once an agent starts searching a lot. So we built the browser version and open-sourced it. How it works: opens a fresh browser on a sticky proxy session and browses a few ordinary pages first then goes to google.com and types the query like a person checks what the page is before parsing, so a captcha never comes back as "0 results" keeps an IP while Google serves it, drops it on the first captcha [If approved:] In our tests, warming the IP up first took the success rate from about 20% to about 80%. You can use it three ways: as an MCP server (Claude Desktop, Cursor, any MCP client), as a local SerpApi-style HTTP API, or as a CLI that writes JSON. It works with any proxy provider that has sticky sessions. We're NodeMaven, a proxy provider, so there's a one-flag setup for ours, and today there's if you want to try it without setting anything up. Honest limits: the first search on a new IP takes a few minutes because of the warm-up, and on Linux Google serves it less often than on Windows - we haven't figured out why yet. If you hit something, tell us here or open an issue.

Google Browser Scraper

Self-hosted Google Search for AI agents, open source

Self-hosted Google Search for AI agents, open source