Puppeteer MCP Server

Overview

The Puppeteer MCP server gives agents a real Chrome browser via Puppeteer: navigate, click, fill, screenshot, evaluate JavaScript. The right pick when the target surface has no API, and a complement (not a replacement) for Playwright MCP for teams that prefer Puppeteer's ergonomics.

Best for

Best for
Teams whose existing tooling is Puppeteer-based, or who need raw browser control

Use cases

• Drive a third-party UI an agent has no API for
• Scrape and parse a complex JavaScript-heavy page
• Capture screenshots for QA or audit trails
• Run scripted browser flows from an agent loop

How to use

Install

npx -y @modelcontextprotocol/server-memory

Tested with

Claude Desktop

Example client config

{\n  "mcpServers": {\n    "puppeteer": {\n      "command": "npx",\n      "args": ["-y", "@modelcontextprotocol/server-puppeteer"]\n    }\n  }\n}

Notes

Why it matters

Most real workflows still hit a surface with no API. The Puppeteer MCP server is one of two cleanest browser-as-tool answers in the reference set.

How teams use it in production

Use for one-off scrapes, third-party UIs without APIs, and visual audit flows. Always wrap in retries and human review for high-stakes runs.

What to watch

Browser-agent surfaces (Browserbase, Browser Use, Puppeteer MCP, Playwright MCP) are converging. Pick on ergonomic fit, not capability gap.