Kagi MCP Server

The Kagi MCP Server exposes two primary tools to MCP-compatible clients via the Kagi API:

• kagi_search_fetch: Performs web, news, videos, podcasts, and image searches. It supports optional page extracts, filters, and Kagi lenses.
• kagi_extract: Fetches the full content of a specific URL and returns it as markdown.

Note: kagi_fastgpt and kagi_summarizer tools have been removed and are planned for future releases.

For local installation, it is recommended to use uv. First, ensure uv is installed on your system:

MacOS/Linux:

curl -LsSf https://astral.sh/uv/install.sh | sh

Windows:

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

Once uv is installed, you can use uvx kagimcp as the command for various MCP clients.

You can use Kagi's hosted MCP server without installing anything locally. Point your HTTP-capable MCP client to https://mcp.kagi.com/mcp and authenticate using your Kagi API key via Bearer HTTP authentication.

Example for Claude Code:

claude mcp add kagi https://mcp.kagi.com/mcp --transport http --header "Authorization: Bearer $(read -sp 'API key: ' k; echo $k)" --scope user

The server can be self-hosted in HTTP mode. In this mode, it is multi-tenant: each request must supply its API key via the Authorization: Bearer <key> header. The provided Dockerfile allows you to run a pinned version of kagimcp from PyPI. The container respects the $PORT environment variable.

Build and run locally:

docker build -t kagimcp-hosted .
docker run --rm -p 8000:8000 kagimcp-hosted

Smoke test the running server:

curl -sL http://127.0.0.1:8000/mcp -X POST \
  -H "authorization: Bearer $KAGI_API_KEY" \
  -H "content-type: application/json" \
  -H "accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Edit your OpenCode configuration file at ~/.config/opencode/opencode.json and add the kagi entry:

{
  "mcp": {
    "kagi": {
      "type": "local",
      "command": ["uvx", "kagimcp"],
      "enabled": true,
      "environment": {
        "KAGI_API_KEY": "<YOUR_API_KEY_HERE>"
      }
    }
  }
}

To use Kagi with Claude Desktop, add the following configuration to your Claude Desktop config file (accessible via Settings -> Developer -> Edit Config):

{
  "mcpServers": {
    "kagi": {
      "command": "uvx",
      "args": ["kagimcp"],
      "env": {
        "KAGI_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Add Kagi to your Kiro MCP config file (~/.kiro/settings/mcp.json for global, or .kiro/settings/mcp.json for project-scoped). Use the same JSON structure as the Claude Desktop configuration:

{
  "mcpServers": {
    "kagi": {
      "command": "uvx",
      "args": ["kagimcp"],
      "env": {
        "KAGI_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Kagi provides a hosted Model Context Protocol (MCP) server that allows AI clients to interact with Kagi's search and data services.

Endpoint: https://mcp.kagi.com/mcp

Authentication: Use Bearer HTTP authentication with an API key obtained from the Kagi API Dashboard. Note that OAuth2 flow is currently not supported but is on the roadmap.

You can integrate the hosted Kagi MCP server into Claude Code using the claude mcp add command. You will need an API key from your Kagi API Dashboard. The server uses Bearer HTTP authentication.

Run the following command to set it up:

claude mcp add kagi https://mcp.kagi.com/mcp --transport http --header "Authorization: Bearer $(read -sp 'API key: ' k; echo $k)" --scope user

You can inspect the server using the Model Context Protocol Inspector.

To inspect the published package:

npx @modelcontextprotocol/inspector uvx kagimcp

To inspect a local checkout:

npx @modelcontextprotocol/inspector uv --directory /ABSOLUTE/PATH/TO/kagimcp run kagimcp

The inspector is typically available at http://localhost:5173.

Install the Kagi MCP server for Claude using Smithery:

npx -y @smithery/cli install kagimcp --client claude