Summary
The ranking is the product; the news feed is just the input.
Anyone can pull a headline feed — the useful part here is sorting it by sentiment with a model rather than a keyword list, which is what stops “not as bad as feared” from being classified as good news. Two keys and a local clone is the whole setup, and both services have free tiers to start on.
What it is
A small MCP server that fetches recent articles from NewsAPI and uses a Cohere model to rank them by positive sentiment, returning the most uplifting stories to your assistant.
What you get
- `fetch_list_of_goodnews` — recent articles ranked by positive sentiment, returned to the client
- Sentiment scoring done zero-shot by an LLM rather than by keyword lists
- Ranking applied across the fresh article set on every request, so the result reflects today rather than a cached list
Requirements
Two API keys — a NewsAPI key and a Cohere key — supplied to the client config as `NEWS_API_KEY` and `COHERE_API_KEY`. Clone the repository and run it with uv, pointing the client at the server directory. Licensed Apache 2.0.