Kagi MCP Server
Model Context Protocol Server for Kagi Search
How to install
With Forge Super installed, the button above opens the app on this extension, where you press Install. Nothing is installed until you do.
You can also open the Extensions panel in the app and search for Kagi MCP Server. The extension's id is kagimcp.
Don't have Forge Super yet? Download it here.
README
From the extension's repository, written by its authors and not by Mentenaz. See the source.
Kagi MCP Server
Kagi Search and Universal Summarizer integration through the Model Context Protocol (MCP). Works with Claude Code, Claude Desktop, Cursor, Zed, and any MCP-compatible client.
This repository provides:
kagiapi- Standalone Rust client library for Kagi's APIs (crates.io)kagi-mcp-server- MCP server binary exposing Kagi tools to AI assistants (crates.io)- Zed extension - Direct integration with the Zed editor via the extension marketplace
Prerequisites
Get Kagi API access:
- Request API access by emailing support@kagi.com (currently in closed beta)
- Get your API key from Kagi Settings
Quick Start
Install via cargo
cargo install kagi-mcp-server
Install via npm (pre-built binaries, no Rust toolchain needed)
npx -y kagi-mcp-server@latest
Zed Editor
- Install the extension: Extensions > Search "Kagi MCP Server" > Install
- Add to your Zed settings:
{
"context_servers": {
"kagimcp": {
"settings": {
"kagi_api_key": "YOUR_KAGI_API_KEY_HERE",
"kagi_summarizer_engine": "cecil"
}
}
}
}
Claude Code (CLI)
claude mcp add -s user kagi -e KAGI_API_KEY="<YOUR-KEY-HERE>" -- kagi-mcp-server
Claude Desktop / Cursor / Other MCP Clients
See the kagi-mcp-server documentation for detailed configuration examples for all supported clients.
Available Tools
| Tool | Description |
|---|---|
kagi_search_fetch | Web search using Kagi's privacy-focused search API |
kagi_fastgpt | Quick AI-powered answers with automatic web search |
kagi_enrich_web | Find non-commercial "small web" content and discussions |
kagi_enrich_news | Find non-mainstream news sources and alternative perspectives |
kagi_summarizer | Summarize web pages, documents, PDFs, and videos |
Configuration Options
| Option | Required | Description |
|---|---|---|
KAGI_API_KEY | Yes | Your Kagi API key |
--summarizer-engine | No | Summarizer engine: cecil (default), agnes, daphne, or muriel |
Usage Examples
Ask your AI assistant:
- "Search for the latest AI safety research"
- "Find recent climate change news"
- "Summarize this article: https://example.com/article"
- "Summarize this YouTube video: https://youtube.com/watch?v=..."
Using the Kagi API Library
If you want to use Kagi's APIs in your own Rust project:
[dependencies]
kagiapi = "0.0.30"
use kagiapi::KagiClient;
#[tokio::main]
async fn main() -> Result<(), kagiapi::Error> {
let client = KagiClient::new("your-api-key");
let results = client.search("rust programming", Some(10)).await?;
Ok(())
}
See the kagiapi documentation for full API reference.
Troubleshooting
API errors?
- Ensure you have Kagi Search API access (closed beta)
- Double-check your API key
Server not responding?
- Test manually:
KAGI_API_KEY=your-key kagi-mcp-server --help - Check your client's logs for error messages
Links
- kagi-mcp-server on crates.io
- kagiapi on crates.io
- NPM Package & Client Configuration
- Kagi Search
- Kagi API Docs
License
MIT License - see LICENSE