MCP Server for the RAG Web Browser Actor 🌐

Implementation of 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 a web search in ChatGPT.

> This MCP server is deprecated in favor of mcp.apify.com
>
> For the same functionality and much more, please use one of these alternatives:

🚀 Recommended: use mcp.apify.com

The easiest way to get the same web browsing capabilities is to use mcp.apify.com with default settings.

Benefits:

✅ No local setup required

✅ Always up-to-date

✅ Access to 6,000+ Apify Actors (including RAG Web Browser)

✅ OAuth support for easy connection

✅ Dynamic tool discovery

Quick Setup:

Go to https://mcp.apify.com

Authorize the client (Claude, VS Code, etc.)

Copy the generated MCP server configuration (or use OAuth flow if supported)

Start using browsing & other tools immediately

🌐 Alternative: direct RAG Web Browser integration

You can also call the RAG Web Browser Actor directly via its HTTP/SSE interface.

Benefits:

✅ Direct integration without mcp.apify.com

✅ Real-time streaming via Server-Sent Events

✅ Full control over the integration

✅ No additional dependencies

Docs: Actor Documentation

---

🎯 What does this MCP server do?

This server is specifically designed to provide fast responses to AI agents and LLMs, allowing them to interact with the web and extract information from web pages.
It runs locally and communicates with the RAG Web Browser Actor in Standby mode,
sending search queries and receiving extracted web content in response.

Web Search: Query Google Search, scrape top N URLs, and return cleaned content as Markdown

Single URL Fetching: Fetch a specific URL and return its content as Markdown

Local MCP Integration: Standard input/output (stdio) communication with AI clients

🧱 Components

Tools

name: search

description: Query Google Search OR fetch a direct URL and return cleaned page contents.
arguments:
- query (string, required): Search keywords or a full URL. Advanced Google operators supported.
- maxResults (number, optional, default: 1): Max organic results to fetch (ignored when query is a URL).
- scrapingTool (string, optional, default: raw-http): One of browser-playwright | raw-http.
- raw-http: Fast (no JS execution) – good for static pages.
- browser-playwright: Handles JS-heavy sites – slower, more robust.
- outputFormats (array of strings, optional, default: [markdown]): One or more of text, markdown, html.
- requestTimeoutSecs (numb

... [truncated — view full README on GitHub]