Playwright MCP

Overview

Playwright MCP wraps Microsoft’s Playwright browser-automation library in an MCP-compatible server, exposing a set of high-level tools (click, type, evaluate, screenshot, etc.) that large-language-model agents can call to explore and interact with web pages.
Instead of screen-scraping pixels, it returns structured accessibility-tree snapshots, making reasoning deterministic, fast, and vision-model-free.

Key Capabilities

• Agent-friendly DOM snapshots – delivers semantic accessibility nodes rather than images.
• Cross-browser support – automate Chromium, Firefox and WebKit head-/headless.
• Deterministic tool APIs – well-scoped commands such as browser_evaluate, page_click, browser_pdf_save.
• Lightweight & fast – minimal overhead by avoiding screenshot-based processing.
• Extensible runtime – enable extra features via --caps flags and deploy via Node.js CLI or Docker.