Summary
Live web search for an agent, with the crawler under your own control.
The single tool is deliberately flexible: an agent can widen or narrow a search by language, region, and domain lists, and steer it toward news or blogs, while the crawler runs on your own infrastructure so the search path is yours rather than a third party's.
What it is
A server that adds a web-search tool to an assistant, backed by a crawler service you run yourself, so the model can look up current information — news, blogs, or general results — with domain and result-type filters.
What you get
- web_search searches the web for a query, with options for the number of results, language and region, domain include/exclude lists, term exclusions, and a result type of all, news, or blogs
Requirements
Runs through npx on Node, and needs the companion crawler service running — provided as a Docker Compose setup — reachable at a URL you configure.
Setup effort
One command — npx -y websearch-mcp