Agent Browser MCP

A Model Context Protocol (MCP) server that provides browser automation capabilities through Vercel’s agent-browser. This enables LLMs to interact with web pages using a fast Rust CLI with Node.js fallback.

Quick Start

# 1. Install agent-browser CLI
npm install -g agent-browser && agent-browser install

# 2. Add to Claude Desktop (or your MCP client)
npx agent-browser-mcp

Then use tools like browser_navigate, browser_click, browser_snapshot to control the browser from your AI agent.

Features

• AI-Optimized Browser Control - Semantic element locators using accessibility properties, text matching, and data attributes
• Session Isolation - Multiple isolated browser sessions with separate cookies, storage, and navigation history
• Comprehensive Automation - Navigation, form filling, clicking, scrolling, keyboard input, and more
• Data Extraction - Get text, HTML, attributes, accessibility snapshots, screenshots, and PDFs
• Cookie Management - Full control over browser cookies and storage
• JavaScript Execution - Run arbitrary scripts in the browser context
• Network Inspection - Monitor console messages and network requests

Use Cases

• Web Scraping for AI - Extract structured data from websites for RAG, research, or analysis
• Automated Testing - AI-powered end-to-end testing with natural language assertions
• Form Automation - Fill forms, submit data, handle multi-step workflows
• Screenshot Analysis - Capture pages for visual AI analysis or documentation
• Session Automation - Login once, persist cookies, automate authenticated workflows
• Data Entry - Bulk data input across web applications
• Monitoring - Track changes on web pages, detect updates, gather metrics

Why Agent Browser?

Agent Browser was built from the ground up for AI agents, not adapted from human-centric automation tools. This fundamental difference delivers significant advantages:

Agent-Native Design

Performance

• Rust-powered CLI - Fast execution with minimal overhead
• Lightweight snapshots - Accessibility tree instead of full DOM
• Streaming output - Real-time feedback without buffering

Installation

npm install agent-browser-mcp

Or run directly with npx:

npx agent-browser-mcp

Prerequisites

• Node.js 18 or newer
• agent-browser CLI installed:

# Install agent-browser globally
npm install -g agent-browser

# Download Chromium browser
agent-browser install

# On Linux, install system dependencies if needed:
# agent-browser install --with-deps

⚠️ Windows Note: agent-browser currently has known issues on Windows with native shells (PowerShell/CMD). For Windows users, we recommend using WSL (Windows Subsystem for Linux) until the upstream issue is resolved.

Configuration

VS Code

Add to your VS Code settings (JSON):

{
  "mcp": {
    "servers": {
      "agent-browser": {
        "command": "npx",
        "args": ["agent-browser-mcp"]
      }
    }
  }
}

Claude Desktop

Add to your claude_desktop_config.json:

Cursor

Add to Cursor MCP settings:

{
  "mcpServers": {
    "agent-browser": {
      "command": "npx",
      "args": ["agent-browser-mcp"]
    }
  }
}

Claude Code

claude mcp add agent-browser -- npx agent-browser-mcp

Custom agent-browser Path

If agent-browser is not in your PATH, specify its location:

{
  "mcpServers": {
    "agent-browser": {
      "command": "npx",
      "args": ["agent-browser-mcp"],
      "env": {
        "AGENT_BROWSER_PATH": "/path/to/agent-browser"
      }
    }
  }
}

Available Tools

Navigation

Interaction

Data Extraction

Element State

Screenshots & PDF

Session Management

Wait Operations

Cookies & Storage

JavaScript & Debugging

Selector Syntax

agent-browser supports semantic locators that are AI-friendly:

# By role and name
button:has-text("Submit")
[role="button"][name="Login"]

# By text content
text=Click here
:has-text("Welcome")

# By accessibility attributes
[aria-label="Search"]
[placeholder="Enter email"]

# By test IDs
[data-testid="submit-button"]

# Standard CSS selectors
#email
.form-input
form > input[type="text"]

Session Management

Create isolated browser sessions for parallel automation:

// Create a new session with custom viewport
const session = await client.callTool({
  name: "browser_new_session",
  arguments: {
    viewport: { width: 1920, height: 1080 }
  }
});

// Use session ID for subsequent operations
const result = await client.callTool({
  name: "browser_navigate",
  arguments: {
    url: "https://example.com",
    sessionId: "session-id-here"
  }
});

// Close when done
await client.callTool({
  name: "browser_close_session",
  arguments: { sessionId: "session-id-here" }
});

Programmatic Usage

import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";

const transport = new StdioClientTransport({
  command: "npx",
  args: ["agent-browser-mcp"],
});

const client = new Client({
  name: "my-browser-client",
  version: "1.0.0",
});

await client.connect(transport);

// Navigate to a page
await client.callTool({
  name: "browser_navigate",
  arguments: { url: "https://example.com" }
});

// Get page content
const text = await client.callTool({
  name: "browser_get_text",
  arguments: {}
});

// Take a screenshot
await client.callTool({
  name: "browser_screenshot",
  arguments: {
    path: "/tmp/screenshot.png",
    fullPage: true
  }
});

Environment Variables

Development

# Clone the repository
git clone https://github.com/minhlucvan/agent-browser-mcp.git
cd agent-browser-mcp

# Install dependencies
npm install

# Build
npm run build

# Run tests
npm test

# Watch mode
npm run dev

# Start server
npm start

Related Projects

• agent-browser - The underlying browser automation CLI by Vercel
• Playwright MCP - Microsoft’s Playwright-based MCP server
• MCP Chrome - Chrome extension-based MCP server
• Model Context Protocol - The MCP specification

License

MIT