Kagi MCP Server

🚀 Kagi Search Proxy Server (Node.js)

A Kagi search proxy server implemented in Node.js, which provides web search functionality for AI assistants by using the Kagi Search API.

🚀 Quick Start

Prerequisites

• Node.js 18 or higher
• Kagi API key

Run without Installation

You can directly run the server using npx:

KAGI_API_KEY=your_api_key npx github:elliottlawson/kagi-mcp-server

Installation Steps

1. Clone the repository:

git clone https://github.com/elliottlawson/kagi-mcp-server.git
cd kagi-mcp-server

2. Run directly (no need to build):

KAGI_API_KEY=your_api_key node build/index.js

If you need to modify the code, you need to install dependencies and rebuild:

npm install
npm run build

Configuration in Claude Desktop

Add the following MCP settings to your Claude Desktop configuration:

{
  "mcpServers": {
    "kagi": {
      "command": "npx",
      "args": ["github:elliottlawson/kagi-mcp-server"],
      "env": {
        "KAGI_API_KEY": "your_api_key"
      }
    }
  }
}

Or if you have cloned the repository:

{
  "mcpServers": {
    "kagi": {
      "command": "node",
      "args": ["ABSOLUTE_PATH_TO_REPO/build/index.js"],
      "env": {
        "KAGI_API_KEY": "your_api_key"
      }
    }
  }
}

Configuration in Cursor

1. Go to Cursor Settings -> MCP -> Add a new MCP server.
2. Configure your MCP:
   • Name: Kagi Search
   • Type: Command
   • Command: npx github:elliottlawson/kagi-mcp-server
   • Environment Variables: KAGI_API_KEY=your_api_key

Ask Questions Requiring Search with an AI Assistant

For example:

How can the Kagi search proxy server help me find online information?

✨ Features

• Conduct web searches using the Kagi API.
• Support multi-threaded queries.
• Format search results.
• Built with TypeScript and the official MCP SDK.
• Pre-built and ready to use (no build steps required).
• A unique tool name (kagi_web_search) to avoid conflicts with other MCP servers.

📦 Installation

The installation steps are included in the "Quick Start" section above.

💻 Usage Examples

Using the Server to Search

You can ask questions that require search using an AI assistant, such as:

How can the Kagi search proxy server help me find online information?

📚 Documentation

Configuration in Different Tools

• Claude Desktop: Add the corresponding MCP settings to the configuration file.
• Cursor: Configure in the Cursor settings.

Debugging Guide

Use the MCP Inspector tool for debugging:

# Install the tool: npm install -g @kagi/mcp-inspector
# Start debugging: mcp-inspector

Add environment variables in the Inspector:

• Click Edit -> Configuration in the top menu bar.
• Add the KAGI_API_KEY variable with the value of your API key.

Development Guide

Project Structure

src/
├── server.ts      # Main server file
└── search.ts     # Search function implementation
build/             # Build output directory

Build and Run

npm run build    # Build the project
npm start        # Run the built product

🔧 Technical Details

• The server communicates through standard input and output.
• The log level can be adjusted via the environment variable LOG_LEVEL, with the default being info.
• The tool name kagi_web_search ensures consistency with the MCP client configuration.

📄 License

No license information is provided in the original document, so this section is skipped.

⚠️ Important Note

First, ensure that you have access to the search API. Currently, this feature is in a closed beta, and you need to apply in advance. Please contact support@kagi.com to get an invitation.

💡 Usage Tip

The log level can be adjusted via the environment variable LOG_LEVEL, with the default being info. Make sure the tool name kagi_web_search is consistent with the MCP client configuration.