Every other search server wants an API key and a billing account; this one wants neither, and the structured title/URL/description shape is exactly what a model needs to decide what to open next. The trade-off is inherent to scraping — no service agreement behind it and a hard ceiling of 10 results — so treat it as a personal-use tool rather than something to put under a product.
A single-tool search server that returns DuckDuckGo results. There is no key to obtain and no account to create: build it, point your client at the binary, and search calls start working.
- `search` takes a `query` string and an optional `limit`
- `limit` defaults to 5 results and accepts up to 10
- Structured results — each one a title, a URL and a description
- No API key or authentication anywhere in the path
Node. Clone the repository, `npm install`, `npm run build`, then run `build/index.js` under node — setup snippets are given for Claude Desktop and for the Claude Dev extension in VSCode. No credentials. Results are scraped from the search page rather than fetched through an API, and the author positions this for personal use. Version 0.1.0.