SearXNG MCP Server

The SearXNG MCP Server bridges the powerful, privacy‑focused search engine SearxNG with AI assistants through the Model Context Protocol. By exposing web‑search and URL‑content extraction as first‑class tools, it lets developers equip conversational agents with up‑to‑date information without compromising user privacy or relying on proprietary APIs. This is especially valuable for teams that need a self‑hosted, open‑source search layer that can be controlled and audited.

At its core the server offers two robust tools. The searxng_web_search tool accepts natural language queries and returns structured search results, supporting pagination, time‑range filtering, language selection, and safe‑search levels. The web_url_read tool fetches a given URL and converts its content into Markdown, with fine‑grained extraction options such as section targeting, paragraph ranges, and heading lists. Together these tools enable agents to browse the web, summarize articles, or pull specific sections from a webpage—all while respecting user‑defined constraints.

Key capabilities include intelligent caching of URL content, which reduces latency and limits repeated requests to external sites. Pagination controls let developers fetch large result sets in manageable chunks, while time‑filtering and language options tailor searches to niche contexts (e.g., news from the last 24 hours in German). Safe‑search levels give an extra layer of content moderation, useful for child‑friendly or corporate environments.

Typical use cases span from building a knowledge‑base assistant that can pull the latest news or academic papers, to creating a data‑collection pipeline that scrapes specific sections from regulatory documents. In research settings, the server can serve as a sandbox for testing search‑driven question answering without exposing sensitive queries to third‑party services. Because the MCP interface is language‑agnostic, any AI client that understands the protocol—Claude, GPT, or custom agents—can invoke these tools seamlessly.

Integrating SearXNG into an AI workflow is straightforward: a developer configures the (and optional authentication or proxy settings), then registers the server with an MCP‑compatible client. The client can call or as part of a larger chain, passing results directly into subsequent reasoning steps. The server’s stateless design and support for environment‑based configuration make it easy to deploy in Docker, Kubernetes, or as a simple Node.js process. Its open‑source nature and reliance on the self‑hosted SearxNG instance give it a distinct advantage over commercial search APIs, offering full control over data residency, privacy policies, and cost.