MCP Browser Use Server

🚀 MCP Server Implementation - Browser-based Usage

This project implements a specific service of the MCP protocol, allowing users to control browsers to perform various operations through natural language instructions. The core functions are as follows:

• Browser Control: Navigation, clicking, inputting, scrolling, etc.
• Form Handling: Filling out forms, submitting requests, etc.
• Status Query: Retrieving page content, taking screenshots, etc.
• Multi-tab Management: Creating, switching, and closing tabs.
• Visual Recognition: Locating elements through image recognition.
• Model Invocation: Supporting multiple AI models for intelligent decision-making.

✨ Features

• Browser Control: Enables navigation, clicking, input, scrolling, and other common browser operations.
• Form Handling: Facilitates form filling and request submission.
• Status Query: Allows users to obtain page content and take screenshots.
• Multi-tab Management: Supports creating, switching, and closing tabs.
• Visual Recognition: Uses image recognition to locate elements on the page.
• Model Invocation: Supports multiple AI models for intelligent decision-making.

📦 Installation

The project relies on the following technologies:

• MCP protocol: Defines the standard interface for AI-service interaction.
• Claude CLI: Provides the runtime environment for the MCP server.
• Python: Used to implement the service logic.
• Uv: Serves as the development toolchain.

🚀 Quick Start

You can easily start the service via the Claude CLI:

# Install and run the service
npx @modelcontextprotocol/cli run mcp-server-browser-use

# View logs
npx @modelcontextprotocol/cli logs mcp-server-browser-use

# Open the console in the Inspector
npx @modelcontextprotocol/inspector run mcp-server-browser-use

📚 Documentation

Configuration Options

You can set the service behavior through configuration files or environment variables. Here are some key configurations:

{
  "command": "uvx",
  "args": ["mcp-server-browser-use"],
  "env": {
    // Browser debugging port, default is 9222
    "CHROME_DEBUGGING_PORT": "9222",
    // Browser user data directory, default is an empty string to use the default configuration
    "CHROME_USER_DATA": "",
    // Whether to maintain the session, default is false
    "CHROME_PERSISTENT_SESSION": false,
    // Model provider, supports anthropic, openai, etc.
    "MCP_MODEL_PROVIDER": "anthropic",
    // Other model parameters...
  }
}

Security Considerations

Since the service needs to control browser behavior, there are certain security risks. It is recommended that you:

1. Use it in a trusted network environment.
2. Avoid deploying it in a production environment.
3. Strictly manage API keys and access permissions.
4. Regularly update dependencies to prevent vulnerabilities.

Development Guide

Environment Setup

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

# Create a virtual environment and install dependencies
python -m venv env
source env/bin/activate  # Use env\Scripts\activate on Windows
pip install -r requirements.txt

Running the Service

uvx run mcp-server-browser-use

Debugging Mode

uvx logs --follow mcp-server-browser-use

💻 Usage Examples

The following browser functions can be implemented through the MCP protocol:

1. Navigate to a specified URL.
2. Simulate user clicks and inputs.
3. Submit form data.
4. Take a snapshot of the page.
5. Manage multiple tabs.
6. Recognize visual elements.

⚠️ Important Note

• Ensure the service is compatible with the MCP protocol.
• Regularly update dependencies to prevent security vulnerabilities.
• Conduct thorough testing before deploying to a production environment.
• Comply with relevant laws, regulations, and user privacy policies.

For more detailed information, please refer to the MCP protocol documentation and Claude CLI documentation.