home / mcp / rag web browser mcp server

RAG Web Browser MCP Server

A MCP Server for the RAG Web Browser Actor

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "apify-mcp-server-rag-web-browser": {
      "url": "https://mcp.apify.com",
      "headers": {
        "APIFY_TOKEN": "YOUR_APIFY_API_TOKEN"
      }
    }
  }
}

This MCP server lets AI agents and LLMs browse the web by querying search engines, fetching web pages, and returning the extracted content in Markdown. It runs locally and communicates with the RAG Web Browser Actor, enabling fast, automated web access for RAG pipelines and agent systems.

How to use

You connect your MCP client to either a local MCP server or a remote endpoint to start browsing and extracting content. The server supports two primary modes: a localstdio MCP server that runs on your machine and a hosted HTTP MCP endpoint you can reach over the network. Use the local stdio mode when you want full control and offline operation, or the hosted endpoint for a managed, always-up connection.

How to install

Prerequisites: Node.js (v18 or higher) and an Apify API token if you plan to use the hosted endpoint.

1) Install the dependencies and prepare the project locally.

2) Run the local MCP server using the included command.

Additional sections

MCP server connections are defined in two ways: a remote HTTP endpoint and a local stdio-based runtime. The HTTP endpoint uses a URL that you connect to with your MCP client. The local stdio runtime runs as a child process and communicates over standard input/output.

Key capabilities include fast web search with optional result scraping, and single URL fetching with content returned as Markdown. The server is designed to provide fast responses to AI agents querying the web and extracting information from web pages.

If you want to run the local MCP server, use the following configuration as a starting point. This shows a local runtime that uses npx to execute the MCP server package and exposes the required environment variable token.

Available tools

search

Query Google Search OR fetch a direct URL and return cleaned page contents. Supports options for max results, scraping method, and output format.