playwright-mcp

Official Microsoft Playwright MCP server, enabling LLMs to interact with web pages using structured accessibility snapshots.

• Source: GitHub Repository
• Category: Testing & Debugging Tools
• Tags: mcp, playwright, browser, automation

Features

• Model Context Protocol (MCP) Server: Provides browser automation using Playwright, optimized for LLM-driven workflows.
• Structured Accessibility Snapshots: Enables interaction with web pages using the accessibility tree, not pixel-based input or screenshots.
• LLM-Friendly: Operates on structured data, removing the need for vision models.
• Deterministic Tool Application: Reduces ambiguity compared to screenshot-based automation.
• Headless and Headed Modes: Supports running the browser in headless mode for background/batch operations or in headed mode (including Linux environments without DISPLAY).
• Client-Server Operation: Can run as a remote server, connecting via WebSocket endpoint.
• Tool Modes:
  • Snapshot Mode (default): Uses accessibility snapshots for performance and reliability.
  • Vision Mode: Uses screenshots and X/Y coordinate-based interactions (with --vision flag).
• Automation Tools (Snapshot Mode):
  • browser_navigate: Navigate to a URL
  • browser_go_back: Go back to previous page
  • browser_go_forward: Go forward to next page
  • browser_click: Click on web page element
  • browser_hover: Hover over element
  • browser_drag: Drag and drop between elements
  • browser_type: Type text into editable element
  • browser_press_key: Press keyboard key
  • browser_snapshot: Capture accessibility snapshot
  • browser_save_as_pdf: Save page as PDF
  • browser_wait: Wait for specified time (up to 10 seconds)
  • browser_close: Close the page
• Automation Tools (Vision Mode):
  • browser_navigate, browser_go_back, browser_go_forward, browser_save_as_pdf, browser_wait, browser_close (as above)
  • browser_screenshot: Capture page screenshot
  • browser_move_mouse: Move mouse to coordinates
  • browser_click: Click at coordinates
  • browser_drag: Drag and drop between coordinates
  • browser_type: Type text at coordinates
  • browser_press_key: Press keyboard key
• Use Cases:
  • Web navigation and form-filling
  • Data extraction from structured content
  • Automated testing driven by LLMs
  • General-purpose browser interaction for agents
• Open Source: Licensed under Apache-2.0

Pricing

This project is open source and available under the Apache-2.0 license. No pricing or paid plans are listed.