ask_fastgptLABSCO SUMMARY
Two different jobs, kept apart: get an answer, or get material.
FastGPT returns a resolved answer; the enrichment endpoints return content for the model to reason over. Keeping them as separate tools means a quick factual check and a research pass do not have to go down the same path.
What it is
An MCP server over the Kagi API, exposing three of its endpoints: FastGPT for answered questions, and the web and news enrichment endpoints for content.
What you get
- A question put to Kagi's FastGPT and answered, rather than returned as a list of links
- Web content pulled in to enrich the model's context on a topic
- Recent news content pulled in the same way, when currency is what the question turns on
- Kagi's own index behind all three, rather than a scrape of somebody else's results
Requirements
A Kagi API key, supplied in the client config; Kagi bills API usage on its own terms. Run it with uv from a local checkout.
Setup effort
Build from source — git clone https://github.com/apridachin/kagi-search-mcp.git
USE CASES
01
Getting a straight answer to a factual question
Put the question to FastGPT and get the answer back, rather than a page of results to read through.
02
Grounding a piece of writing in current sources
Enrich the context with web content on the topic, so what gets written is based on retrieved material rather than recall.
03
Catching up on something that is still developing
Pull recent news content when the answer depends on what happened this week rather than what is generally true.
04
Checking a claim before you repeat it
Ask FastGPT, then pull the web content behind the topic to see whether the answer holds up.
3 tools
Listed in project docs. Your client may show a different list.
enrich_webenrich_newsBEFORE YOU INSTALL
What to know first
Tools you get
3 tools named in the project's documentation. Fastgpt, enrich/web and enrich/news are resources wrapping the same API methods, not three more tools. All three tools read.
Requires
- uv on PATH
- KAGI_API_KEY in your environment
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
This one is built from source rather than installed with a single command.
Server setup
TERMINAL
git clone https://github.com/apridachin/kagi-search-mcp.git
cd kagi-search-mcp
uv syncMCP Clients
Claude Desktop
CLAUDE DESKTOP · JSON
{
"mcpServers": {
"kagi-mcp": {
"command": "uv",
"args": [
"--directory",
"<YOUR_PATH_TO_PROJECT>",
"run",
"kagi-mcp"
],
"env": {
"KAGI_API_KEY": "<YOUR_KAGI_API_KEY>"
}
}
}
}<YOUR_PATH_TO_PROJECT> is the full path of your kagi-search-mcp clone (for example /Users/you/kagi-search-mcp).
