Playwright MCP

Playwright MCP is a Model Context Protocol server built by Microsoft that enables LLMs to interact with web pages through structured accessibility snapshots. Rather than relying on screenshots or vision models, it operates on the page's accessibility tree for fast, deterministic browser automation that works with any MCP-compatible client.

Playwright MCP bridges the gap between AI agents and real browsers, giving coding assistants and autonomous workflows the ability to navigate, click, type, and extract data from web pages through a standardized protocol.

• Accessibility-Based Interaction - Uses Playwright's accessibility tree instead of pixel-based input, enabling fast and deterministic element targeting without vision models
• Core Browser Automation - Provides tools for clicking, typing, navigating, hovering, dragging, form filling, file uploads, dialog handling, and keyboard input
• Page Snapshots - Captures structured accessibility snapshots of pages that are more useful to LLMs than raw screenshots
• Tab Management - List, create, close, and switch between browser tabs within a single session
• Code Execution - Run arbitrary Playwright code snippets and evaluate JavaScript expressions directly on pages or elements
• Screenshot and PDF Generation - Take viewport, element, or full-page screenshots and save pages as PDFs with optional capabilities
• Network and Console Monitoring - Inspect network requests and console messages for debugging and observability
• Persistent and Isolated Profiles - Run with a persistent browser profile that retains login state across sessions, or use isolated mode for clean testing contexts
• Browser Extension Mode - Connect to existing browser tabs using the Playwright MCP Bridge Chrome extension to leverage already logged-in sessions
• Configuration File Support - Configure browser type, viewport, proxy, timeouts, allowed origins, secrets, and more through a JSON config file or command-line arguments
• Docker Support - Run headless Chromium automation inside a Docker container with the official Microsoft container image
• Multi-Client Compatibility - Works with VS Code, Cursor, Windsurf, Claude Desktop, Claude Code, Goose, Cline, Codex, Copilot, Amp, LM Studio, Warp, and many more MCP clients
• Vision and DevTools Capabilities - Opt into coordinate-based mouse interactions and browser developer tools via capability flags
• Test Assertions - Generate locators and verify element visibility, text presence, and input values for test automation workflows
• Code Generation - Automatically generates TypeScript code for browser interactions performed through the MCP tools