Web Scout Mcp

Web Scout MCP does two things and does them without ceremony. First, DuckDuckGo web search is exposed as an MCP tool you can call from any MCP-compatible client — Claude Desktop, a custom host, or Pink Pixel's own Navi MCP client. Second, URL text extraction uses Cheerio to strip a page down to readable text a model can consume. Search is tunable: region, time range, and a max-results cap let you keep result volume sane instead of dumping fifty links into the context window. The implementation is TypeScript/Node.js, installed from npm, and stateless — you can run it locally or remotely. There's no API key, registration, or subscription, which is the whole point: for a hobbyist or a researcher prototyping an LLM research assistant on a zero budget, the friction is close to nil. Where it hurts is precisely where the free endpoint helps. DuckDuckGo's underlying search is rate-limited and can block aggressive request patterns, so throughput is unpredictable and you have no SLA to point at. There's no headless browser, so JavaScript-rendered pages and single-page apps come back as a shell of empty content — a big deal if your targets are modern web apps. Extraction is plain text only; you don't get full HTML or structured data out of it, so schema-aware scraping is out. And it's self-host only — no official cloud version — so a team that wants zero ops overhead will need to stand it up themselves. The honest read: this is a great default for personal research assistants, MCP prototyping, and coding assistants that occasionally need a current page. It is not a substitute for Tavily or SerpAPI at production volume, and it is not a scraper. Pair it with a paid search API when your query volume goes up, and keep it around for the zero-cost work that doesn't need guarantees.