Summary
Search and read, with nothing to sign up for.
Most search servers want a key before they answer anything. This one queries DuckDuckGo and then fetches the page itself, so a question goes from query to page text in two calls — and the same code, self-hosted, has no rate limit at all.
What it is
Two tools: a DuckDuckGo search returning titles, URLs and snippets, and a fetcher that strips a page down to its readable content.
What you get
- Up to 10 DuckDuckGo results per query, 5 by default
- Readable text from any URL, with navigation, ads and footers removed
- No signup step — the search goes to DuckDuckGo directly
- A built-in cache, 5 minutes on the free tier
Requirements
Node. Run it from a clone, with npx -y github:tiagohanna123/web-search-mcp, or from the ghcr.io image. The free tier allows 10 requests a minute and extracts 5K characters per page; a managed instance is a one-off lifetime purchase, and the MIT-licensed code self-hosted has neither limit.
Setup effort
One command — npx -y github:tiagohanna123/web-search-mcp