BBEANPICK See the product
Independent specialty coffee discovery

Good coffee,
better discovered.

BeanPick brings specialty coffees from independent roasteries into one searchable place. Explore by origin, variety, process and tasting notes, compare pack sizes, then go straight to the roaster’s own store.

Explore the productHow it works

A discovery and monitoring tool, not a resale marketplace. The roaster owns the sale.

Illustrative mockup · sample values
460clean catalog records
21roasteries in the snapshot
23map reference countries; 21 listed
From scattered listings to a clearer choice

Coffee details should help you choose.

One catalog makes it easier to compare the details that matter, while keeping each roastery’s official store as the source for the final purchase.

01

Search beyond the label

Find coffees by roaster, origin, variety, process and tasting notes, with filters for decaf and current offers.

Search & filters
02

Compare the bag fairly

See package sizes and price per 100 g when the roaster listing provides enough information.

Unit-price comparison
03

Follow origin and variety

Explore a coffee-origin map and a separate variety guide, then open matching coffees from the catalog.

Origin atlas
04

Keep an eye on change

The desktop app can detect new listings, price changes, sell-outs and restocks. Saved coffees can trigger a restock notification.

Desktop monitoring
Built around the roaster, not around a checkout

The coffee stays with its source.

BeanPick organizes public product details for discovery. When you are ready, product links take you to the roaster’s own store. BeanPick does not sell or resell coffee.

01
Collect and normalizeDesktop collection adapters bring supported roastery listings into a consistent product record.
02
Check the dataQuality rules flag or exclude uncertain records, and a prior snapshot can be retained when a source fails.
03
Link back to the roasterThe catalog helps with discovery; availability and checkout remain on the official store.
Snapshot, not live inventory. The iPhone web app displays the latest published data. Prices and availability can change; confirm them on the roaster’s site before buying.
A real product, with clear limits

What the current build does today.

The October 8, 2026 published snapshot contains 460 clean product records from 21 roasteries. The public iPhone web app reads that published snapshot; collection and scheduled rechecks run in the Electron desktop app when it is open.

Published dataset460 records · Oct 8, 2026
Desktop recheck intervalEvery 2 hours while running
FavoritesSaved locally on this device
PurchasesAt the roaster’s official store
A measured AI roadmap

Useful AI should show its work.

The current code already has model-assisted image-note extraction and OCR paths. Claude is not integrated today; the ideas below are planned evaluation work, not shipped features.

In current code

Image notes with source evidence

The collection pipeline includes a Gemini image-model path and local OCR fallback. Tasting-note evidence stores the source text or source URL so extracted notes can be traced back to a roastery listing.

Planned · Claude API

Grounded extraction and search

Evaluate Claude for structured extraction from inconsistent roastery pages, conflict flags for human review, and translating a shopper’s natural-language request into catalog filters. Results would still come only from the current product dataset.

Technical guardrails: keep Claude credentials out of the static website; validate model output against a strict schema; attach source evidence; cache repeat work; cap cost and latency; and never let a model invent stock, prices or products. The existing public web app is a static snapshot, so any Claude API call would require a controlled server-side or desktop integration.

Discover with context. Buy from the source.

BeanPick is a founder-led specialty-coffee discovery project in active development.

Back to top