an1604/news-mcp-server

GenAI News Assistant MCP Server

This repository contains a template for building a Model Context Protocol (MCP) server that provides tools for fetching and analyzing news information using NewsAPI.org and a Large Language Model.

Project Structure

.
├── python/                # Python implementation
│   ├── Dockerfile         # Python Docker configuration
│   └── src/               # Python source code
├── typescript/            # TypeScript implementation
│   ├── Dockerfile         # TypeScript Docker configuration
│   └── src/               # TypeScript source code
└── README.md              # This file

Prerequisites

• Docker installed on your system
• NewsAPI.org API key (register for free at https://newsapi.org/)
• OpenAI API key (for LLM functionality)

Environment Variables

The following environment variables are required:

• NEWS_API_KEY: Your NewsAPI.org API key
• OPENAI_API_KEY: Your OpenAI API key

Building and Running

Python Version

# Build the Docker image
cd python
docker build -t news-assistant-python .

# Run the container
docker run -p 3000:3000 \
  -e NEWS_API_KEY=your_news_api_key \
  -e LLM_API_KEY=your_llm_api_key \
  news-assistant-python

TypeScript Version

# Build the Docker image
cd typescript
docker build -t news-assistant-typescript .

# Run the container
docker run -p 3000:3000 \
  -e NEWS_API_KEY=your_news_api_key \
  -e LLM_API_KEY=your_llm_api_key \
  news-assistant-typescript

Testing

You can validate that the server works as intended in development using the MCP Inspector https://github.com/modelcontextprotocol/inspector

MCP Tools

The server should implement the following MCP tools:

1. search_news: Search for recent news articles matching a specific query
2. extract_information_from_article: Extract structured information from a news article
3. extract_key_info_and_sentiment: Analyze news articles for key entities and sentiment

For detailed specifications of each tool, please refer to the assignment documentation.

Development

Choose either the Python or TypeScript implementation based on your preference. Both implementations provide the same functionality through the MCP interface.

License

This is a private repository for take-home assignment purposes only.