Puppeteer MCP

open sourcestdio

Browser automation for agents: navigate, click, screenshot, and scrape the web

Updated 2026-02-28

BrowserFree· Open source, self-hosted. Puppeteer downloads Chromium on first run.GitHub Docs

Official MCP server that gives AI agents full browser automation via Puppeteer. Agents can navigate to URLs, interact with elements, fill forms, take screenshots, and execute JavaScript, enabling web scraping, UI testing, and browser-based automation without a headless browser setup.

Install

npx -y @modelcontextprotocol/server-puppeteer

Runtime: node

Tools (7)

puppeteer_navigate

Navigate to a URL in the browser

urllaunchOptionsallowDangerous

puppeteer_screenshot

Take a screenshot of the current page or a specific element

nameselectorwidthheightencoded

puppeteer_click

Click an element on the page by CSS selector

selector

puppeteer_hover

Hover over an element on the page

selector

puppeteer_fill

Fill out an input field

selectorvalue

puppeteer_select

Select an option in a dropdown

selectorvalue

puppeteer_evaluate

Execute JavaScript in the browser console and return the result

script

Use Cases

·Scrape websites that require JavaScript rendering
·Automate form filling and multi-step web workflows
·Take screenshots for visual monitoring or documentation
·Test web applications end-to-end with an AI agent
·Extract data from paywalled or dynamic content pages

Key Points

•Full Puppeteer browser control via MCP, navigate, click, fill, screenshot, run JS
•Official Anthropic-maintained server
•Screenshots returned as base64-encoded images, usable by vision-capable models
•Supports custom launch options for proxy, viewport, and headless configuration

Do-Nothing Score

Find out how close you are to Ghost CEO.

Take the quiz

Machine-Readable

This MCP is available as structured JSON via the API.

GET /api/v1/mcps/puppeteer