Kagi MCP server
Setup Intructions
Before anything, unless you are just using non-search tools, ensure you have access to the search API. It is currently in closed beta and available upon request. Please reach out to [email protected] for an invite.
Install uv first.
MacOS/Linux:
curl -LsSf https://astral.sh/uv/install.sh | sh
Windows:
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
Installing via Smithery
Alternatively, you can install Kagi for Claude Desktop via Smithery:
npx -y @smithery/cli install kagimcp --client claude
Setup with OpenAI
Codex CLI
To add the Kagi mcp server to codex cli, you will need to use the following command:
codex mcp add kagi --env KAGI_API_KEY=<YOUR_API_KEY_HERE> -- uvx kagimcp
This will write out the configuration to ~/.codex/config.toml, so if you need to update/rotate your API key, update your key there before running codex again.
Codex CLI comes with its own built-in search (via --search flag), but it is disabled by default. So to deconflict between search and Kagi, just don't enable it.
Setup with Claude
Claude Desktop
// claude_desktop_config.json
// Can find location through:
// Hamburger Menu -> File -> Settings -> Developer -> Edit Config
{
"mcpServers": {
"kagi": {
"command": "uvx",
"args": ["kagimcp"],
"env": {
"KAGI_API_KEY": "YOUR_API_KEY_HERE",
"KAGI_SUMMARIZER_ENGINE": "YOUR_ENGINE_CHOICE_HERE" // Defaults to "cecil" engine if env var not present
}
}
}
}
Claude Code
Add the Kagi mcp server with the following command (setting summarizer engine optional):
claude mcp add kagi -e KAGI_API_KEY="YOUR_API_KEY_HERE" KAGI_SUMMARIZER_ENGINE="YOUR_ENGINE_CHOICE_HERE" -- uvx kagimcp
Now claude code can use the Kagi mcp server. However, claude code comes with its own web search functionality by default, which may conflict with Kagi. You can disable claude's web search functionality with the following in your claude code settings file (~/.claude/settings.json):
{
"permissions": {
"deny": [
"WebSearch"
]
}
}
Pose query that requires use of a tool
e.g. "Who was time's 2024 person of the year?" for search, or "summarize this video: https://www.youtube.com/watch?v=jNQXAC9IVRw" for summarizer.
Debugging
Run:
npx @modelcontextprotocol/inspector uvx kagimcp
Local/Dev Setup Instructions
Clone repo
git clone https://github.com/kagisearch/kagimcp.git
Install dependencies
Install uv first.
MacOS/Linux:
curl -LsSf https://astral.sh/uv/install.sh | sh
Windows:
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
Then install MCP server dependencies:
cd kagimcp
# Create virtual environment and activate it
uv venv
source .venv/bin/activate # MacOS/Linux
# OR
.venv/Scripts/activate # Windows
# Install dependencies
uv sync
Setup with Claude Desktop
Using MCP CLI SDK
# `pip install mcp[cli]` if you haven't
mcp install /ABSOLUTE/PATH/TO/PARENT/FOLDER/kagimcp/src/kagimcp/server.py -v "KAGI_API_KEY=API_KEY_HERE"
Manually
# claude_desktop_config.json
# Can find location through:
# Hamburger Menu -> File -> Settings -> Developer -> Edit Config
{
"mcpServers": {
"kagi": {
"command": "uv",
"args": [
"--directory",
"/ABSOLUTE/PATH/TO/PARENT/FOLDER/kagimcp",
"run",
"kagimcp"
],
"env": {
"KAGI_API_KEY": "YOUR_API_KEY_HERE",
"KAGI_SUMMARIZER_ENGINE": "YOUR_ENGINE_CHOICE_HERE" // Defaults to "cecil" engine if env var not present
}
}
}
}
Pose query that requires use of a tool
e.g. "Who was time's 2024 person of the year?" for search, or "summarize this video: https://www.youtube.com/watch?v=jNQXAC9IVRw" for summarizer.
Debugging
Run:
# If mcp cli installed (`pip install mcp[cli]`)
mcp dev /ABSOLUTE/PATH/TO/PARENT/FOLDER/kagimcp/src/kagimcp/server.py
# If not
npx @modelcontextprotocol/inspector \\
uv \\
--directory /ABSOLUTE/PATH/TO/PARENT/FOLDER/kagimcp \\
run \\
kagimcp
Then access MCP Inspector at http://localhost:5173. You may need to add your Kagi API key in the environment variables in the inspector under KAGI_API_KEY.
Advanced Configuration
- Level of logging is adjustable through the
FASTMCP_LOG_LEVELenvironment variable (e.g.FASTMCP_LOG_LEVEL="ERROR")- Relevant issue: https://github.com/kagisearch/kagimcp/issues/4
- Summarizer engine can be customized using the
KAGI_SUMMARIZER_ENGINEenvironment variable (e.g.KAGI_SUMMARIZER_ENGINE="daphne")- Learn about the different summarization engines here
- The
--httpcli option can be used to toggle streamable HTTP transport on. Can use along with--portand--hostargs.
Recommend MCP Servers 💡
voice-mode
VoiceMode MCP brings natural conversations to Claude Code
@mindpilot/mcp
Visualize code architecture and flows for AI agents via MCP server with local processing.
nerve-hq/nerve-mcp-server
Implements an MCP server to integrate with the Nerve API, enabling AI to access and process customer feedback and other data from Nerve.
coresignal_data_api
The Coresignal MCP Server integrates Coresignal's B2B data (companies, employees, job postings) with AI applications, allowing AI assistants to access fresh data directly within chat interfaces.
apollographql/apollo-mcp-server
Apollo MCP Server enables AI assistants like Claude Desktop to interact with any GraphQL API by exposing GraphQL operations as AI-accessible tools, facilitating real-time data queries.
shadow-cljs-mcp
An MCP server that monitors shadow-cljs builds and provides real-time build status updates, allowing LLMs to verify build status after making changes to ClojureScript files.