MCP/Browser MCP
Browser MCP logo

Browser MCP

Verified Blue (500+ GitHub stars)

Control, inspect, and automate your local Chrome browser with AI using fast, private, native automation.

frontend•Browser Automation & Testing
GitHub890
License:MIT
VERIFIED
ByBrowser MCP@Browser MCP

About this Server

Browser MCP is an open Model Context Protocol server and Chrome extension pairing that enables AI coding assistants like Claude Code, Cursor, Windsurf, and VS Code to automate web browsers directly on your local machine. Because automation runs locally through your real Chrome profile, all active logins and cookies are preserved, latency is near zero, and no activity or private session tokens are sent to remote third-party cloud servers.

Real-time Interaction

Direct low-latency RPC protocol communication with structured JSON schemas.

Scoped Sandboxing

Restricted capability bounds to prevent accidental modifications or leakage.

Security & Best Practices

Local Execution Sandbox: All browser communications run through localhost RPC with zero cloud telemetry or DOM recording relay.
Session Privacy: Uses your existing local Chrome profile so cookies and auth tokens remain solely on your physical device.
No Token Exfiltration: Credentials and passwords entered during automation are never transmitted to external third-party servers.
Visual Transparency: All AI actions execute visibly in your local browser window, allowing real-time monitoring and immediate cancellation.

When to Use (and When Not To)

Best For
  • •Local web development verification, frontend UI testing, and smoke test workflows directly from your editor
  • •Automating complex end-to-end user flows that require active authenticated sessions without creating fake mock logins
  • •Inspecting dynamic Single Page Applications (SPAs) and JavaScript-rendered DOM states with zero network round-trip delay
  • •Capturing visual regression screenshots and console errors during live coding sessions
Not For / Limitations
  • •Headless cloud CI/CD pipelines lacking a desktop display server or Chrome browser instance
  • •Massive distributed web scraping across thousands of parallel worker nodes
  • •Enterprise corporate environments where installing local browser extensions is strictly prohibited by security group policy

Install & Setup

terminal command
$claude mcp add browsermcp -- npx @browsermcp/mcp@latest

Exposed Tools & Endpoints

10 tools

Functions your AI agent can invoke via this Model Context Protocol server

navigate

Navigate active browser tab to any specified URL or local development server port.

click

Click interactive DOM elements targeting CSS selectors, text matches, or viewport coordinates.

type

Type text into input fields, textareas, and rich-text editors with simulated keystrokes.

screenshot

Capture high-resolution viewport or full-scroll page screenshots for visual verification.

snapshot

Extract accessibility tree hierarchies and structured DOM states for semantic LLM reasoning.

get_console_logs

Retrieve browser runtime error logs, console warnings, and uncaught JavaScript exceptions.

wait

Pause execution until a specified element appears, navigation completes, or timeout expires.

hover

Hover cursor over navigation menus, dropdowns, and interactive tooltips to reveal UI states.

press_key

Dispatch keyboard key events such as Enter, Escape, Arrow keys, or keyboard shortcuts.

drag_drop

Execute drag and drop operations between source and target DOM coordinates.