SearxNG

You run a small MCP server that lets Large Language Models search the web privately via SearxNG. It’s designed to be lightweight and easy to integrate with MCP-compatible clients, so you can add web search capabilities to your AI workflows without exposing user data.

How to use

You connect to the MCP server from your MCP client by using one of the supported runtime methods. The server runs locally (stdio transport) or can be accessed via an HTTP-based transport if you choose the SSE path in a different setup. For typical uses, start the server through the recommended runtime (pipx, uvx, Python module, or Docker) and point your client to the local endpoint or the configured URL. The MCP client will send search queries and receive structured results optimized for model consumption.

How to install

Prerequisites you need before starting:

• Python 3.10 or higher

• A SearxNG instance (public or self-hosted)

Using pipx run (Recommended, no installation required)

{
  "mcpServers": {
    "searxng": {
      "command": "pipx",
      "args": [
        "run", "searxng-simple-mcp@latest"
      ],
      "env": {
        "SEARXNG_MCP_SEARXNG_URL": "https://your-instance.example.com"
      }
    }
  }
}

Using uvx run (No installation required)

{
  "mcpServers": {
    "searxng": {
      "command": "uvx",
      "args": [
        "run", "searxng-simple-mcp@latest"
      ],
      "env": {
        "SEARXNG_MCP_SEARXNG_URL": "https://your-instance.example.com"
      }
    }
  }
}

Using Python with pip (requires installation)

{
  "mcpServers": {
    "searxng": {
      "command": "python",
      "args": ["-m", "searxng_simple_mcp.server"],
      "env": {
        "SEARXNG_MCP_SEARXNG_URL": "https://your-instance.example.com"
      }
    }
  }
}

Using with Docker (No installation required)

{
  "mcpServers": {
    "searxng": {
      "command": "docker",
      "args": [
        "run", "--rm", "-i", "--network=host",
        "-e", "SEARXNG_MCP_SEARXNG_URL=http://localhost:8080",
        "ghcr.io/sacode/searxng-simple-mcp:latest"
      ]
    }
  }
}

Notes about Docker usage

When using Docker with MCP servers, pass environment variables directly with -e within the args array, since the env object may not propagate automatically to the container.

Transport protocols

The MCP server supports STDIO (default) for CLI-style integrations and SSE for HTTP-based clients. STDIO is suitable for Claude Desktop and similar MCP-compatible clients, while SSE starts an HTTP server for web-based interactions. Not all applications support SSE, so verify compatibility before using it.

Configuration

Configure the server using environment variables. The following options are available:

• SEARXNG_MCP_SEARXNG_URL: URL of the SearxNG instance to use • Default: https://paulgo.io/

• SEARXNG_MCP_TIMEOUT: HTTP request timeout in seconds • Default: 10

• SEARXNG_MCP_DEFAULT_RESULT_COUNT: Default number of results to return • Default: 10

• SEARXNG_MCP_DEFAULT_LANGUAGE: Language code for results (e.g., 'en', 'ru', 'all') • Default: all

• SEARXNG_MCP_DEFAULT_FORMAT: Default format for results ('text', 'json') • Default: text

• SEARXNG_MCP_LOG_LEVEL: Logging level (e.g., 'DEBUG', 'INFO', 'WARNING', 'ERROR', 'CRITICAL') • Default: ERROR

• TRANSPORT_PROTOCOL: Transport protocol ('stdio' or 'sse') • Default: stdio

Security and notes

Increase log levels with care; DEBUG/INFO may produce excessive output and interfere with some MCP clients.

Troubleshooting

If you cannot reach your SearxNG instance, verify SEARXNG_MCP_SEARXNG_URL is correct and reachable from the running MCP server. For Docker setups, ensure network access matches your host configuration and, if needed, use --network=host to connect to a host-bound SearxNG service.

Development

To test locally, install dependencies and run the server module directly. Use the appropriate transport depending on your client.

Publishing to PyPI

Package maintainers can publish new versions following the standard Python packaging workflow. Ensure dependencies are installed and use the usual publishing tools.