Kagi Search — Claude Code MCP server: what it does, how to install
Kagi Search connects Kagi web search and page content extraction to your AI client. Supplies fresh web results and converts URLs to markdown.
no license updated 2026-10-29
Open on GitHubWhat Kagi Search does
The server equips assistants with two tools: kagi_search_fetch executes queries across web pages, news feeds, videos, and podcasts with custom lenses, while kagi_extract fetches target pages directly as markdown. Earlier summarizing and FastGPT helpers were dropped, so synthesis now relies on your active model.
Operation requires a valid KAGI_API_KEY. You can connect to Kagi's managed HTTP endpoint at https://mcp.kagi.com/mcp using bearer authentication or run the package locally via uvx. Without the uv package manager installed, local execution fails immediately.
What it helps with
- run web and news queries using Kagi search indexes
- pull full web page content directly as clean markdown
- filter search results with domain lists and Kagi lenses
Example prompts for the agent
- Search the web for the latest updates on Python 3.13
- Find recent podcast episodes discussing distributed databases
How to install Kagi Search
Install
The MCP registry has no launch command for it: see the repository README
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Kagi Search
How to install Kagi Search in Claude Code?
Run claude mcp add kagi -e KAGI_API_KEY="YOUR_API_KEY_HERE" -- uvx kagimcp in your terminal. You can also connect directly to the hosted server at https://mcp.kagi.com/mcp using HTTP transport.
What does Kagi Search do?
It provides tools named kagi_search_fetch and kagi_extract to MCP clients. The assistant uses them to query web resources and retrieve page contents as markdown text.
What is required to run the server locally?
You need an active Kagi API key set in KAGI_API_KEY and the uv tool installed on your system.