Web Search MCP, developed by Mrkrsl, is a free MCP server built with TypeScript and Node.js. It enhances AI clients' ability to retrieve real-time information from the web. Distributed as a GitHub release zip, this tool integrates with various search engines to provide essential web search capabilities. It is designed to extend the reach of AI agents and local language models by offering direct access to current online data, making it a valuable addition for advanced AI applications.

What does Web Search MCP connect to?

The Web Search MCP server establishes connections with several popular web search engines, including Bing, Brave, and DuckDuckGo. This enables it to offer varied web search capabilities. As an MCP (Model Context Protocol) server, its core function is to extend the real-time information retrieval abilities of AI clients. It acts as a bridge, allowing AI systems to access and process current data directly from the internet, enriching their operational context.

Core features and tools

This server exposes specific callable tools designed for web interaction. These tools are named exactly as documented: full-web-search, for performing detailed web queries; get-web-search-summaries, for quickly obtaining search results and condensed information; and get-single-web-page-content, for fetching and extracting content from a specified URL. These functions provide AI clients with precise methods to interact with and gather data from the internet.

Who is it for?

The primary audience for Web Search MCP includes developers and users working with Local LLMs and AI agents who require real-time access to web data. It serves those who need to equip their AI systems with the ability to perform searches, gather information, and extract content directly from the internet. This tool is compatible with AI clients and platforms such as Local LLMs, LM Studio, and LibreChat. Skip this if your AI client does not support MCP servers or local execution.

How to get started

No credentials are required. The server communicates with AI clients via stdio transport, supporting platforms like Local LLMs, LM Studio, and LibreChat for local execution. Configuration involves pointing the AI client's MCP settings, such as in "mcp.json" or "librechat.yaml", to the server's "dist/index.js" file after local setup.

Use cases

• Performing extensive web searches and extracting full page content for in-depth research, enabling AI agents to gather detailed information on any topic.
• Obtaining quick search results and summaries for efficient information gathering, allowing AI clients to rapidly grasp key points from web queries.
• Fetching and extracting content from a specific URL, which is useful for targeted data retrieval from known online sources.

Web Search MCP provides a direct pathway for AI clients to access current web information without needing API keys. It integrates with multiple search engines and offers distinct tools for search and content extraction. This makes it a practical solution for enhancing AI agent capabilities.