Summary

Ten calls a day is enough to find out whether it works on your pages.

No signup, no key, no card — connect and try it on the specific sites that have been defeating your fetch tool. That is the right way to evaluate a scraper, because the only number that matters is whether it renders the pages you actually need. If it does, the free account triples the allowance before you decide about credits.

What it is

A hosted scraping and search service. Pages are fetched by real Chrome browsers with JavaScript executed, then converted to Markdown an LLM can read — which is what makes single-page apps, Cloudflare-protected pages and government sites work rather than returning an empty shell.

What you get

`scrape` converts any URL to clean, LLM-optimised Markdown, and handles SPAs, dynamic content and PDFs
`crawl` runs a Google search for a query and scrapes the top results to Markdown in one call
`search` returns Google results as structured JSON with titles, URLs and snippets
A stated 84% success rate including JavaScript-heavy sites, Cloudflare-protected pages and government sites
The same operations are available as an HTTP API with per-call pricing in USDC — `POST /scrape` at $0.003, `POST /crawl` at $0.005, `POST /serp/search` at $0.002
Discovery endpoints for agents: an agent card, an OpenAPI spec and a health check

Requirements

Nothing to install and no API key to start: point any client at `https://anybrowse.dev/mcp` over Streamable HTTP. Published as anybrowse version 1.0.4, with the registry entry at dev.anybrowse/anybrowse version 1.0.0. The free tier is 10 MCP calls per day per IP anonymously, or 50 per day with a free account; beyond that, credits from $5 or per-call payment in USDC on Base through the x402 protocol.

Setup effort

One command — npx -y mcp-remote https://anybrowse.dev/mcp