Kagi MCP Server
by elliottlawson
A Node.js implementation of the Kagi MCP server that provides web search capabilities to AI assistants using the Kagi search API. It's pre-built and ready to use, requiring no build step.
Last updated: N/A
What is Kagi MCP Server?
The Kagi MCP Server is a Node.js application that acts as a Model Context Protocol (MCP) server, enabling AI assistants to perform web searches using the Kagi Search API.
How to use Kagi MCP Server?
To use the server, you need a Kagi API key. You can run it directly using npx or clone the repository and run it with Node.js. Configuration examples are provided for Claude Desktop and Cursor, specifying the command, arguments, and environment variables.
Key features of Kagi MCP Server
Web search using the Kagi API
Support for multiple search queries in parallel
Formatted search results
Built with TypeScript and the official MCP SDK
Pre-built and ready to use
Unique tool name (
kagi_web_search)
Use cases of Kagi MCP Server
Integrating web search into AI assistants
Providing AI assistants with up-to-date information
Enabling AI assistants to answer questions requiring web search
Avoiding conflicts with other search tools in MCP environments
FAQ from Kagi MCP Server
What is the Kagi API?
What is the Kagi API?
The Kagi API is a search API that provides web search capabilities.
Do I need a Kagi API key to use this server?
Do I need a Kagi API key to use this server?
Yes, a Kagi API key is required. Access to the search API is currently in closed beta and available upon request from [email protected].
Do I need to build the project before running it?
Do I need to build the project before running it?
No, the repository includes pre-built JavaScript files, so no build step is required to use it.
How do I configure this server with Claude Desktop?
How do I configure this server with Claude Desktop?
Add the provided MCP config to your Claude Desktop configuration, specifying the command, arguments, and environment variables.
How do I debug the server?
How do I debug the server?
Run the MCP Inspector using the provided command and access it at http://localhost:5173. You may need to add your Kagi API key in the environment variables in the inspector.