A coffee expert in your Vellum assistant. Finds great beans and equipment matched to your taste, budget, and kitchen — and gets sharper over time as it learns what you actually like.
No credentials, no external API keys. Curated origin and equipment knowledge is baked in; live pricing and availability come from Vellum's built-in web search.
What it does
- Learns your taste. A persistent profile stores your preferred roasts, flavors, brew methods, budget, kitchen constraints, and equipment. Every bean you rate teaches it: 4-5 stars adds an origin to your favorites, 1-2 stars steers future picks away from it.
- Recommends beans. Scores 15 curated origins against your profile and hands back optimized web-search queries so the assistant surfaces current, buyable bags with live prices — not just origin names.
- Recommends equipment. Ranks 18 curated espresso machines, grinders, and brewers across entry → commercial tiers, matched to your budget, experience, and brew method.
- Teaches when you're curious. Two skills carry deep reference material on processing methods, roast levels, terroir, grinder types, and the upgrade ladder — pulled in only when you want to understand the why.
Surfaces
| Surface | Kind | What it does |
|---|---|---|
coffee_profile | tool | View/update your taste profile; log beans and gear with auto-learning |
coffee_find_beans | tool | Rank origins by match score and generate live-search queries |
coffee_find_machine | tool | Rank equipment across price tiers to your budget and experience |
bean-discovery | skill | Bean recommendation workflow + coffee education |
equipment-guide | skill | Equipment selection workflow + gear education |
Install
assistant plugins install coffee-aficionado # once whitelisted in the marketplace
assistant plugins install https://github.com/vellum-ai/coffee-aficionado # direct, pre-marketplace
Usage
Just talk to your assistant about coffee:
- "What beans should I try next? I like bright, fruity light roasts."
- "I loved that Ethiopian Yirgacheffe — 5 stars." → logged, and Ethiopia is now a learned favorite.
- "I've got $600 for an espresso setup, total beginner." → matched machine + grinder picks with live prices.
The more you log, the better it gets.
Data & privacy
Your taste profile lives in data/taste-profile.json inside the plugin's data
directory, on your machine. Nothing is sent anywhere except the web searches the
assistant runs to find current beans and prices.
License
MIT