Model Context Protocol (MCP) server that exposes DuckDuckGo search as a tool for LLM clients (Claude Desktop, Cursor, VS Code, ...), backed by a self-hosted SearXNG metasearch instance.

SearXNG aggregates results from DuckDuckGo, Google, Brave, Wikipedia, and more with its own request handling — no browser fingerprinting, no API keys, fully self-hosted.

Part of the MCP Search Servers project:

Repo	What it is
google-mcp	Same design for Google search

Service

Service	Port	MCP tool	Description
`duckduckgo-mcp`	3002	`duckduckgo_search`	DuckDuckGo results via SearXNG
`searxng`	—	—	Internal metasearch engine (not published)

The server exposes:

/sse + /messages/ — MCP SSE transport
/search?q=...&num=N — plain HTTP JSON endpoint
/health — liveness probe

Quick Start

cp .env.example .env        # fill in SEARXNG_SECRET
docker compose up -d --build

Generate a SearXNG secret:

python3 -c "import secrets; print(secrets.token_hex(32))"

Verify:

curl http://localhost:3002/health
curl "http://localhost:3002/search?q=python&num=3"

Then point your LLM client at http://localhost:3002/sse.

Tool Contract

{
  "name": "duckduckgo_search",
  "arguments": { "query": "Python programming language", "num_results": 5 }
}

Returns numbered results with title, URL, and snippet (max 20 results, queries capped at 500 characters).

Project Structure

.
├── duckduckgo-mcp/    # MCP server (SSE transport + tools)
│   ├── server.py
│   └── Dockerfile
├── lib/               # Shared search engine implementations
│   ├── duckduckgo_search.py
│   ├── google_search.py  # also used by the google-mcp sibling repo
│   ├── playwright_manager.py
│   └── rate_limiter.py
├── searxng-settings.yml
├── test_client.py
└── docker-compose.yml