Facilitates multi-query web searches and dynamic page scraping, optimizing results for LLMs using TOON encoding.
Sponsored
searXNG-mcp is a FastMCP-based server designed to provide powerful web search and scraping capabilities, specifically optimized for integration with Large Language Models (LLMs). It leverages a running SearXNG instance for multi-query web searches and offers dual scraping methods—fast HTTP requests for static content and full browser rendering with Playwright for dynamic JavaScript-heavy pages. A key innovation is its use of TOON encoding/decoding, which ensures token-efficient communication, making the retrieved data more suitable for LLM processing. The server boasts a modular architecture, robust retry logic, comprehensive logging, and configurable environment variables, making it a reliable tool for augmenting AI agents with real-time web data.
Key Features
010 GitHub stars
02Multi-query web search via SearXNG with configurable result counts
03Dual scraping methods (requests for static, browser for dynamic JavaScript pages)
04TOON encoding for token-efficient LLM communication
05Persistent browser instance reuse for better performance
06Retry logic with exponential backoff for robust operation
Use Cases
01Integrating advanced web data retrieval into applications requiring efficient LLM communication
02Collecting content from dynamic, JavaScript-rendered websites for data analysis or AI training
03Empowering LLMs and AI agents with real-time, up-to-date web search results
