Summary
A search result you can read, not just a list of links.
Searching and reading sit behind one connection, so an assistant can take the results that look right and pull their text without a second integration. The extractor accepts an array of URLs, which makes the follow-up read one call rather than one per link — and since there is no key to obtain, the whole setup is a config entry.
What it is
Two tools: a DuckDuckGo search that returns a structured list of findings, and an extractor that pulls the readable text out of a page — or of several pages passed as an array.
What you get
- A DuckDuckGo search over keywords, a question or a topic, returning titles, URLs and snippets, ten results by default and as many as you ask for
- The readable content of a webpage extracted from its URL — scripts, styles and navigation removed — or of a list of URLs fetched in parallel
- Request throttling and memory limits in front of both, so a batch of pages does not take the process down with it
Requirements
Node.js 18.0.0 or newer, and nothing else: no account and no key. Apache-2.0.
Setup effort
One command — npx -y @pinkpixel/web-scout-mcp@latest