Apify MCP Server Rag Web Browser

🚀 Model Context Protocol (MCP) Server for RAG Web Browser Actor 🌐

This project implements an MCP server for the RAG Web Browser Actor. This Actor serves as a web browser for large language models (LLMs) and RAG pipelines, similar to the web search feature in ChatGPT.

🔄 What is the Model Context Protocol?

The Model Context Protocol (MCP) enables AI applications (and AI agents), such as Claude Desktop, to connect to external tools and data sources. MCP is an open protocol that allows for secure and controlled interactions between AI applications, AI agents, and local or remote resources.

🎯 Features of this MCP Server

The RAG Web Browser Actor allows AI assistants to:

• Perform web searches on Google, scrape the top N URLs from the results, and return their cleaned content in Markdown format.
• Scrape a single URL and return its content in Markdown format.

🧱 Components

Tools

• search: Search for a query term or URL on Google, scrape the top N URLs from the results, and return the cleaned text content.
  • Parameters:
    • query (string, required): The search term or URL.
    • max_results (number, optional): The maximum number of search results to scrape (default: 1).

Prompts

• search: Search for a query phrase or URL on Google and return the scraped web page content in text or Markdown format.
  • Parameters:
    • query (string, required): The search term or URL.
    • max_results (number, optional): The maximum number of search results to scrape (default: 1).

Resources

This server does not provide any resources or prompts.

🛠️ Configuration

Prerequisites

• MacOS or Windows
• The latest version of Claude Desktop (or other MCP clients) installed
• Node.js (v18 or higher)
• Apify API Token (APIFY_API_TOKEN)

Installation

Claude Desktop

Configure Claude Desktop to recognize the MCP server.

1. Open your Claude Desktop configuration file and edit the following files:

   • On MacOS: ~/Library/Application\ Support/Claude/claude_desktop_config.json
   • On Windows: %APPDATA%/Claude/claude_desktop_config.json

   "mcpServers": {
     "mcp-server-rag-web-browser": {
       "command": "npx",
       "args": [
         "/path/to/mcp-server-rag-web-browser/build/index.js"
       ],
       "env": {
         "APIFY-API-TOKEN": "your-apify-api-token"
       }
     }
   }
   

2. Restart Claude Desktop

   • Completely exit Claude Desktop (make sure it is not minimized or closed).
   • Restart Claude Desktop.
   • Check the 🔌 icon to confirm that the Exa server is connected.

3. Examples You can ask Claude to perform web searches, for example:

   What is an MCP server and how can it be used?
   What is an LLM, and what are the recent news updates?
   Find and analyze recent research papers about LLMs.
   

👷🏼 Development

Local Development

If you are working on an unpublished server, you can access the local server through the following command:

"mcpServers": {
    "mcp-server-rag-web-browser": {
      "command": "/path/to/mcp-server-rag-web-browser/build/index.js"
    },
    "env": {
        "APIFY-API-TOKEN": "your-apify-api-token"
    }
}

Local Client

To test the server, you can use example_client:

node build/example_client.js

This script will start the MCP server, get the available tools, and then call the search tool with a query term.

Debugging

Call the RAG Web Browser Actor for testing:

APIFY_API_TOKEN=your-apify-api-token node build/example_call_web_browser.js

Since the MCP server runs through standard input/output (stdio), debugging can be challenging. For the best debugging experience, use the MCP Debugger.