mcp-searxng

What is mcp-searxng?

mcp-searxng is an MCP (Model Context Protocol) server that enables an AI agent to search for new external information through the open-source SearXNG meta-search engine. It integrates SearXNG with Microsoft's markdownify to extract web page content into Markdown-formatted text, targeting developers who need privacy-controlled, multi-engine search for AI agents.

How to use mcp-searxng?

Clone the repository, then run the service using Docker Compose (recommended) or start it manually via uv run server.py --searxng_url=<url>. For Docker, set the SEARXNG_URL environment variable and expose port 5488. The server exposes an SSE endpoint at http://<host>:<port>/sse; configure your MCP host with a URL entry like "url": "http://localhost:5488/sse".

Key features of mcp-searxng

• SSE‑based transport for remote, decoupled operation.
• SearXNG meta‑search engine aggregates results from multiple engines.
• Two tools: web_search (query) and web_url_read (URL to Markdown).
• Self‑hosted or community‑hosted SearXNG compatible.
• Cloud‑native, process‑decoupled architecture.

Use cases of mcp-searxng

• AI agent searches for real‑time, externally‑sourced information.
• Privacy‑sensitive environments requiring self‑hosted search infrastructure.
• Aggregating results from DuckDuckGo, Ecosia, Brave, and others via one endpoint.
• Extracting web pages into clean Markdown for further processing.

FAQ from mcp-searxng

What is SearXNG and why use it with mcp-searxng?

SearXNG is an open‑source meta‑search engine that queries multiple search engines (Google, DuckDuckGo, Ecosia, Brave, etc.) from a single interface. It gives AI agents broader, privacy‑controlled search results without relying on a single engine.

What transport does mcp-searxng use?

It uses Server‑Sent Events (SSE) for transport, allowing the server and client to be decoupled processes (potentially on different nodes). This is more suitable for cloud‑native deployments than STDIO‑based MCP servers.

What are the runtime dependencies?

The project requires Python (managed via uv), Docker for the SearXNG service, and optionally Node.js for the MCP Inspector testing tool. The SearXNG service runs in a Docker container.

Where does search data live?

SearXNG is self‑hosted (either via the provided docker-compose.yaml or your own instance). No data is sent to third‑party services; network requests are made from your server to the upstream search engines.

How do I configure the SearXNG endpoint?

Set the --searxng_url command‑line argument or the SEARXNG_URL environment variable to point to your SearXNG instance (default: http://localhost:8888).