Let Your AI Browse the Web and Scrape Pages with MCP

intermediateSetup time: 15 minutes

Two MCP servers cover web access: Fetch (for simple URL fetching and content extraction) and Puppeteer (for full browser control, including JavaScript-heavy pages, screenshots, and form interaction). Together they make your AI a capable web agent.

Servers You'll Need

Step-by-Step Setup

  1. 1

    Choose your server based on need

    Use Fetch for simple pages and documentation. Use Puppeteer when the page requires JavaScript to render, needs a login, or you want screenshots.

  2. 2

    Install Fetch

    `npm install -g @modelcontextprotocol/server-fetch`

  3. 3

    Install Puppeteer server

    `npm install -g @modelcontextprotocol/server-puppeteer`. Puppeteer downloads Chromium automatically on first run.

  4. 4

    Add both to your config

    Add both servers to your AI client config. They work independently — use whichever is more appropriate for a given task.

Example Prompts to Try

  • Fetch the content of this documentation page and summarize the API
  • Take a screenshot of example.com and describe what you see
  • Scrape the pricing table from this competitor's website
  • Check if this URL returns a 200 status code
  • Fill in and submit this form on the test site

Tips

  • Fetch is faster and lighter — prefer it for static content.
  • For Puppeteer tasks, be specific about what you want captured (text, screenshot, specific element).
  • Web scraping may violate a site's terms of service — check before automating.

FAQ

Can the AI log into a site using Puppeteer?

Yes, Puppeteer can fill forms and click buttons. You'd tell the AI the credentials (or have them in a config). Be careful with this — only use it on sites you own or have permission to automate.

Does Fetch work with paywalled content?

No. Fetch makes an unauthenticated HTTP request. You'd need Puppeteer with a logged-in session for paywalled content.