master-fetch

Why we included this project

Hound is an MCP server that gives AI agents like Claude Code, Cursor, or OpenCode real web access without a paid subscription or a pile of API keys. Everything runs locally in one process: it routes each request to the right tool, handling simple fetches, crawls, searches, and even scanned PDFs, and uses stealth techniques to get past Cloudflare-style bot walls that block most scrapers. Pages come back cleaned with Trafilatura so the model reads the article instead of the site's boilerplate, plus structured signals like content age, page type, and relevance that let an agent act on facts rather than parse raw HTML. You can also plug in your own search keys for Serper, Tavily, or similar services when you want more control. For teams that need live external grounding and would rather self-host than pay a scraping service, it is a practical fit.