Kagi Search

Add to your client

Copy the config for your MCP client and paste it into its config file.

{
  "mcpServers": {
    "kagi-search": {
      "command": "uvx",
      "args": [
        "kagimcp"
      ],
      "env": {
        "KAGI_API_KEY": "<your-kagi-api-key>"
      }
    }
  }
}

Before you start

• The uv package manager installed (uvx is used to launch the server)
• Python (installed/managed automatically by uv)
• A Kagi API key generated from the Kagi API portal (Generate Key in API Management), set as KAGI_API_KEY
• A Kagi account with a payment method on file — the Search API is pay-per-use, not free

About Kagi Search

Tools & capabilities (2)

When to use it

• Use it when you want a privacy-first, ad-free search source feeding your AI assistant.
• Use it when you need to search across multiple media types (web, news, videos, podcasts, images) from one tool.
• Use it when you want full page content returned as clean markdown for the model to read (kagi_extract).
• Use it when you want to apply Kagi lenses or filters to constrain search results.
• Use it when you are already a Kagi subscriber and want to reuse your search quality in agent workflows.

Quick setup

1. 1Install the uv package manager using the platform install script.
2. 2Generate a Kagi API key from the Kagi API portal (API Management → Generate Key).
3. 3In Claude Desktop, open Settings → Developer → Edit Config and add a `kagi` server with command `uvx`, args `["kagimcp"]`, and env `KAGI_API_KEY`.
4. 4Restart the MCP client so it launches the server via uvx.
5. 5Verify by asking the assistant to run a Kagi search and confirm kagi_search_fetch returns results.

Kagi Search FAQ