Because the items are cached locally and filterable, "what happened in my feeds today" and "has anyone written about X this week" are both one query — over the sources you chose rather than whatever a search engine surfaces.
A feed reader with a local SQLite cache in front of it. It fetches and stores your subscriptions, filters them by category, source or keyword, and returns the recent items as structured records — title, summary, publication time, origin and link — rather than as pages to parse.
- Fetching and caching of RSS feeds into a local SQLite database, so repeat questions do not re-fetch every source
- Filtering by category, source or keyword
- Feed management — add, update and delete the feeds in the collection
- Article fetching through Firecrawl, for when the feed's summary is not the whole story
Bun as the runtime, and a Firecrawl API key in `FIRECRAWL_API_KEY` for the article-fetching path. `bun install`, then run `apps/mcp-server/src/mcp-cli.ts` from your client config with `cwd` set to the repository. `PORT` defaults to 5556 and `DB_DIR` decides where the database lives, defaulting to `~/.mcp-rss-crawler`. Set `DEBUG=mcp-rss:*` when you need to see what a failing fetch is doing.