Browser Use Server

🚀 Browser Use Server

This is an MCP (Middleware Communication Protocol) server for web browser automation. It supports various functions, including web page screenshotting, HTML retrieval, JavaScript execution, and console log retrieval, and allows multi-step operations through commands.

🚀 Quick Start

This server is designed for web browser automation. You can perform various operations on web pages, such as taking screenshots, retrieving HTML, executing JavaScript, and getting console logs.

✨ Features

1. Screenshot

• Parameters:
  • url: Required, the URL of the target web page.
  • full: Optional, default is false. Set to true to capture a full-page screenshot.
  • steps: Optional, comma-separated operation steps or instructions.

2. Get HTML

• Parameters:
  • url: Required, the URL of the target web page.
  • steps: Optional, comma-separated operation steps or instructions.

3. Execute JavaScript

• Parameters:
  • url: Required, the URL of the target web page.
  • script: Required, the JavaScript code to be executed.
  • steps: Optional, comma-separated operation steps or instructions.

4. Get Console Logs

• Parameters:
  • url: Required, the URL of the target web page.
  • steps: Optional, comma-separated operation steps or instructions.

📦 Installation

Prerequisites

1. Install Xvfb (for headless browser environment):

   sudo apt-get install xvfb
   

2. Install Miniconda or Anaconda, and create a Conda environment:

   conda create -n browser-use python=3.8
   conda activate browser-use
   

Install via Smithy

1. Clone the repository:

   git clone https://github.com/browser-use/server.git
   cd server
   

2. Install dependencies and build the server:

   npm install
   npm run build
   

📚 Documentation

MCP Configuration

{
  "browser-use": {
    "llm-providers": [
      "glhf", // Use the deepseek-ai/DeepSeek-V3 model
      "ollama", // Use the qwen2.5:32b-instruct-q4_K_M model with a context length of 32k
      "groq", // Use the deepseek-r1-distill-llama-70b model
      "openai", // Use the gpt-4o-mini model
      "openrouter", // Use the deepseek/deepseek-chat model
      "github", // Use the gpt-4o-mini model
      "deepseek", // Use the deepseek-chat model
      "gemini" // Use the gemini-2.0-flash-exp model
    ],
    "vision-support": false, // Visual features are disabled by default. Enable it to improve performance; set to true to enable visual support
    "use-xvfb": true // Automatically detect and use Xvfb to run the environment variable RUNNING_UNDER_XVFB
  }
}

Usage Instructions

Run the Server

npm start

Example Usage

Example 1: Modify Web Page Elements

Change the text color on the page to red:
browser.use.steps(
  "Open the browser",
  "Visit the URL http://example.com",
  "Find all paragraph elements and set the color to red"
)

Configuration Options

• LLM Providers: Support multiple large language models, specifically:
  • glhf: Use the deepseek-ai/DeepSeek-V3 model
  • ollama: Use the qwen2.5:32b-instruct-q4_K_M model with a context length of 32k
  • groq: Use the deepseek-r1-distill-llama-70b model
  • openai: Use the gpt-4o-mini model
  • openrouter: Use the deepseek/deepseek-chat model
  • github: Use the gpt-4o-mini model
  • deepseek: Use the deepseek-chat model
  • gemini: Use the gemini-2.0-flash-exp model

Error Handling

The server will output detailed error information in the console, which is convenient for troubleshooting.

Debugging Tools

npm run debug

Use Cases

• Web page automation testing
• Data scraping and analysis
• Dynamic web page interaction operations
• Multi-step task execution

📄 License

MIT License