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

This repository implements an MCP server for the RAG Web Browser Actor, enabling LLMs and RAG pipelines to interact with the web. It facilitates web search and content extraction, functioning similarly to web search in ChatGPT.

This server provides fast responses to AI agents, running locally and communicating with the RAG Web Browser Actor in Standby mode. It sends search queries and receives extracted web content. The RAG Web Browser Actor allows AI assistants to perform web searches, scrape URLs, and fetch single URLs, returning content as Markdown.

Key features include the search tool, which queries Google Search, scrapes URLs, and returns cleaned content in Markdown. It accepts arguments like query, maxResults, scrapingTool ('browser-playwright' or 'raw-http'), outputFormats ('text', 'markdown', 'html'), and requestTimeoutSecs. The server integrates with AI Agents, enabling web searches and browsing. It operates over stdio and requires Node.js, Claude Desktop, and an Apify API token.