kimcp
MCP server enabling LLMs to use Korean APIs (Naver, Kakao, etc.)
claude mcp add --transport stdio zeikar-kimcp python main.py \ --env SK_APP_KEY="your_sk_app_key" \ --env NAVER_CLIENT_ID="your_naver_client_id" \ --env KAKAO_REST_API_KEY="your_kakao_rest_api_key" \ --env NAVER_CLIENT_SECRET="your_naver_client_secret"
How to use
KiMCP (Kiorea-integrated Model Context Protocol) is an MCP server that exposes several 국내 API integrations (Naver, Kakao, SK Open API/TMAP) through the MCP framework so you can query them from an LLM-powered workflow. The server uses the uv Python environment to run and relies on environment variables for API keys; when a key is not provided the corresponding tool remains disabled, allowing you to enable only the services you need. Typical usage involves starting the server locally with the required API keys and then invoking MCP-enabled prompts or tooling that reference KiMCP endpoints to perform searches (e.g., Naver blog/news/cafe, Kakao maps, SK Open API) and routing results back to your application.
How to install
Prerequisites:
- Python (and a working Python environment)
- uv (Astral uv) installed and accessible in your PATH
- API keys for Naver, Kakao, and SK Open API if you want to enable corresponding tools
Installation steps:
- Clone the repository and navigate into it
git clone https://github.com/zeikar/kimcp
cd kimcp
- Install Python dependencies via uv (as described in the project):
uv sync
- Create the environment file from the example and populate keys:
# Copy the example env to a real env
cp .env.example .env
Edit .env to include:
NAVER_CLIENT_ID=your_naver_client_id
NAVER_CLIENT_SECRET=your_naver_client_secret
KAKAO_REST_API_KEY=your_kakao_rest_api_key
SK_APP_KEY=your_sk_app_key
- Run the MCP server (via Claude Desktop integration or directly with uv):
uv run mcp install main.py -f .env
# or to start development inspector
uv run mcp dev main.py
- If changes are made, restart Claude Desktop as needed to apply updates.
Additional notes
Tips and caveats:
- API keys are optional per service; if you omit a key, the corresponding MCP tool is automatically disabled. You can enable only the services you need by providing the relevant keys.
- The server is designed to be used with Claude Desktop via the MCP workflow; ensure Claude Desktop can access the running process.
- If you experience missing endpoints, verify that the .env file is correctly referenced when starting the server and that the keys are valid.
- Environment variable names to look for: NAVER_CLIENT_ID, NAVER_CLIENT_SECRET, KAKAO_REST_API_KEY, SK_APP_KEY.
- This setup assumes you are using the Python/uv path described in the README. If you switch to a different runner, adjust the mcp_config accordingly (e.g., npm or docker-based deployments).
- For debugging, use the development inspector: uv run mcp dev main.py
Related MCP Servers
MCP-Bridge
A middleware to provide an openAI compatible endpoint that can call MCP tools
mcp-pinecone
Model Context Protocol server to allow for reading and writing from Pinecone. Rudimentary RAG
asterisk
Asterisk Model Context Protocol (MCP) server.
rlm-claude
Recursive Language Models for Claude Code - Infinite memory solution inspired by MIT CSAIL paper
web-research-assistant
MCP server for SearXNG with 13 production-ready tools for web search, package info, GitHub integration, error translation, API docs, and more
llm-bridge
A model-agnostic Message Control Protocol (MCP) server that enables seamless integration with various Large Language Models (LLMs) like GPT, DeepSeek, Claude, and more.