SearXNG Mul MCP

SearXNG Mul MCP is an MCP server that connects MCP clients like Claude Desktop to the privacy-focused SearXNG metasearch engine.

It directly handles search requests through the Model Context Protocol, which enables your AI assistants to directly access real-time web results without compromising your privacy.

Features

• 🌐 Multi-Query Parallel Search: Run several search queries at once instead of waiting for each to finish sequentially
• 🔌 Dual Transport Support: Works with both stdio (for direct MCP client connections) and HTTP protocols
• ⚙️ SearXNG API Integration: Communicates directly with SearXNG’s REST API without browser automation
• 🔒 Basic Authentication: Connects to protected SearXNG instances requiring username/password
• 🐳 Docker Deployment: Ready-to-run container images for easy deployment
• 🛠️ Environment Configuration: Customize behavior through simple environment variables

Features List

• ⚡ Multi-Query Parallel Search: Run a batch of search queries simultaneously instead of one by one.
• 🔌 Dual Transport Support: Offers both stdio for local clients and HTTP for web or networked integrations.
• 🔒 Basic Authentication: Works with private SearXNG instances that are protected by Basic Auth.
• 🐳 Docker Ready: Comes with Docker and Docker Compose files for a quick and isolated deployment.
• ⚙️ Environment Configuration: All settings are managed through simple environment variables for easy setup.
• 🎯 Direct API Integration: Communicates directly with the SearXNG search API for reliable performance.

Use Cases

• AI Research Automation: An AI agent can investigate a topic from multiple angles at once. Instead of searching for “topic history,” then “topic applications,” then “topic controversies” sequentially, it can execute all queries in a single, parallel request to gather comprehensive data much faster.
• Private Data Enrichment for Internal Tools: If you run a private, self-hosted SearXNG instance with curated search engines, you can use this server to let your internal AI tools access that specific information without touching the public web. The Basic Auth support keeps it secure.
• Building Hybrid AI Applications: You could have a core AI service that uses the server’s HTTP endpoint for its web front-end, while a companion desktop application for power users connects to the same server instance via the stdio protocol. This provides flexibility without needing separate backends.

How To Use It

1. For a direct connection with an MCP client (the default stdio mode), run this in your terminal:

SEARXNG_URL=https://your.searxng.com npx -y searxng-mul-mcp

2. If you need to access it over the network, start it in HTTP mode:

SEARXNG_URL=https://your.searxng.com npx -y searxng-mul-mcp --transport=http --host=0.0.0.0 --port=3000

3. Configuration is handled through environment variables.

• SEARXNG_URL: The URL of your public or private SearXNG instance. (Required)
• USERNAME: The username for Basic Authentication. (Optional)
• PASSWORD: The password for Basic Authentication. (Optional)
• TRANSPORT: Set to stdio or http. (Optional)
• HOST: The host address for HTTP mode, typically 0.0.0.0. (Optional)
• PORT: The port for HTTP mode, like 3000. (Optional)
• DEBUG: Set to true to enable detailed logging. (Optional)

4. To integrate it with Claude Desktop, you edit the claude_desktop_config.json file. Add the following server configuration:

{
  "mcpServers": {
    "searxng-mul-mcp": {
      "command": "npx",
      "args": ["-y", "searxng-mul-mcp"],
      "env": {
        "SEARXNG_URL": "https://your.searxng.com",
        "USERNAME": "your_username",
        "PASSWORD": "your_password"
      }
    }
  }
}

5. For a more permanent setup, Docker is the way to go. Create a docker-compose.yml file:

services:
  searxng-mul-mcp:
    image: ghcr.io/jae-jae/searxng-mul-mcp:latest
    environment:
      - SEARXNG_URL=https://your.searxng.com
      # - USERNAME=your_username
      # - PASSWORD=your_password
    ports:
      - "3000:3000"

Then, launch it with docker-compose up -d.

6. The server exposes a single tool named search. It accepts several parameters to refine the search:

• queries: An array of search strings. This is the only required parameter.
• engines: A list of specific search engines to use (e.g., “google”, “bing”).
• categories: Filter results by categories like “general” or “news”.
• safesearch: Set the safe search level (0 for off, 1 for moderate, 2 for strict).
• language: Specify a language code, such as “en” for English.

FAQs

Q: What’s the main benefit of this server compared to other search tools?
A: The key advantage is parallel query execution. Most search tools handle one query at a time, but this one can process a whole batch at once, which dramatically speeds up information gathering for AI agents.

Q: Do I have to run my own SearXNG server?
A: No, you can point it to a public SearXNG instance. However, for privacy, reliability, and customization, deploying your own SearXNG server is the recommended approach.

Q: What happens if my SearXNG instance is down?
A: The server has built-in error handling, including an automatic retry mechanism for network errors and clear messages for authentication failures. If the connection fails persistently, you should check your SEARXNG_URL and ensure the instance is running correctly.