Puppeteer MCP Server

What is the Puppeteer MCP Server?

The Puppeteer MCP Server bridges language models and real web browsers. By managing a headless Chrome/Chromium instance via Puppeteer, this server allows AI agents to navigate web pages, click buttons, fill forms, take visual screenshots, and inspect JavaScript console logs dynamically.

Available Browser Automation Tools

• puppeteer_navigate: Open any HTTP/HTTPS URL in an active headless browser tab.
• puppeteer_screenshot: Capture full-page or element-level screenshots returned directly to the agent.
• puppeteer_click: Click buttons, navigation tabs, or interactive links using CSS selectors.
• puppeteer_fill: Populate input fields and form elements.
• puppeteer_evaluate: Run arbitrary JavaScript in the page context to inspect DOM state.

Installation & Config

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

Frequently Asked Questions (FAQ)

Can the Puppeteer MCP server handle single-page apps (SPAs)?

Yes. Because it uses a real Chromium browser engine, React, Vue, and Angular applications execute client-side JavaScript completely before the agent inspects the page.

Does it support taking visual screenshots for multimodal models?

Yes, captured screenshots can be passed directly to vision-capable models like Claude 3.5 Sonnet or GPT-4o for visual UI debugging.

Can Puppeteer MCP bypass basic authentication?

Yes, custom launch arguments and credentials can be supplied through environment variables.