Puppeteer MCP

Puppeteer MCP is the official Model Context Protocol server for browser automation, maintained as part of the MCP reference server collection. It enables AI assistants to control a real Chromium browser, navigating pages, clicking elements, filling forms, taking screenshots, and executing JavaScript, all through structured MCP tool calls. The server provides seven core tools: puppeteer_navigate for loading URLs with configurable timeouts, puppeteer_screenshot for capturing full-page or element-specific screenshots, puppeteer_click and puppeteer_hover for element interaction via CSS selectors, puppeteer_fill for populating input fields, puppeteer_select for dropdown menus, and puppeteer_evaluate for running arbitrary JavaScript in the browser context. Console logs are captured and accessible for debugging. Deployment supports local execution via npx, Docker with headless Chromium, and Cloudflare Workers for cloud environments. Security controls include an allowDangerous parameter that blocks unsafe launch options like --no-sandbox by default. Screenshots and console logs are stored as MCP resources that AI assistants can reference in subsequent interactions. The server is published as @modelcontextprotocol/server-puppeteer on npm and works with Claude Desktop, Claude Code, Cursor, and other MCP-compatible clients.