browser-scraper MCP Server

About browser-scraper

browser-scraper is an MIT-licensed Node.js library, CLI and MCP server. It drives your local Chrome over the raw DevTools Protocol and never sends Runtime.enable, one of the most widely fingerprinted automation signals. No cloud, no API key, no per-page billing.

Add it to Claude Code:
claude mcp add browser-scraper -- npx -y @eurafaeldecarvalho/browser-scraper mcp

Tools: navigate, snapshot, click, type, press_key, go_back, wait, markdown, metadata, screenshot, solve_turnstile, scrape, check_egress.

- snapshot returns an accessibility outline with [ref=…] ids; it reaches closed shadow roots and cross-origin iframes.
- markdown returns the main content as clean Markdown, with an optional query to keep only relevant blocks.
- scrape escalates from plain HTTP to the browser only when needed.

Guard rails: --allowed-domains refuses navigation outside a list; --secrets .env lets the agent type {{PASSWORD}} placeholders that are filled at keystroke time and redacted from output; page content comes back inside <untrusted-page-content> markers.

Limits: it is not a CAPTCHA solver and not a guaranteed bypass. Against hard targets, IP reputation matters more than the fingerprint. Use it on sites where you are allowed to automate.

Requires Node.js 20+ and a local Chrome/Chromium. I'm the author.

How to connect browser-scraper to Claude, Cursor or VS Code

browser-scraper is self-hosted. Install it from the source repository, then register its start command in your MCP client's config (Claude Code, Cursor, Windsurf or VS Code).

{
  "mcpServers": {
    "browser-scraper": {
      "command": "<start command from the README>",
      "args": []
    }
  }
}

The exact command (often npx, uvx or docker) is in the project README at github.com/eurafaeldecarvalho/browser-scraper.