How it works

From a roastery's website to a traceable farm on the map.

Open Coffee Hub runs a pipeline of six steps, automatically and on a regular cycle, to keep its database of specialty coffee farms, harvests and roasteries fresh and connected. Here's exactly what happens behind the scenes.

1.Curate
We curate the roasters ourselves: quality over quantity.
Open Coffee Hub isn't a list of every coffee shop on earth. We keep a hand-picked roster of the world's finest specialty roasters (the ones known for origin-focused, traceable coffee) and concentrate on covering each of them properly.
No pay-to-list, no dragnet. And any roaster can put themselves forward.
2.Read
An AI confirms each one is a genuine specialty roaster.
Before we invest in reading a whole catalogue, we visit the site and ask an AI: "Is this a specialty coffee roastery?" The AI reads the page, looks for signs of origin-focused roasting (single-farm sourcing, harvest names and periods) and gives a yes or no. Only confirmed roasters move forward.
This step keeps junk out of the database. Chains, tea shops, and supermarket cafes are quietly rejected.
3.Extract
We read every product listing and extract the coffee's story.
For each confirmed roastery, we read every coffee listing and an LLM pulls out the facts that matter, to the depth the roaster actually published: the origin (a named farm, producer, washing station, mill or cooperative, or only a region or country), the process, variety, harvest, altitude, the roaster's cupping score, tasting notes, price, availability and images. Every fact is grounded against the roaster's own words: we keep the verbatim source quote, and anything we can't find in their text is marked unverified rather than shown as fact.
No invented data. If a roaster named only a country, the origin stays country-level, and every fact carries a "how sure we are" marker: stated by the roaster · derived from a name/region · unverified.
4.Match
We connect coffees from different roasters that share the same farm harvest.
This is the core of Open Coffee Hub. When two different roasters both source from the same farm and the same harvest, we link their products to a single harvest record. We compute a "mapping signature" from the farm name, country, varietals, processing method, and harvest period. Products that share a signature are siblings: different roasters, same green coffee.
That's how you can discover that the Ethiopia Guji harvest from Roastery A and the "Yirgacheffe Natural" from Roastery B are actually the same coffee, priced differently and roasted differently.
5.Check
We show how sure we are, never a made-up grade.
Rather than invent a quality score, we surface honest, checkable signals: a per-field confidence marker (stated by the roaster · derived · unverified) and cross-roaster corroboration (when several roasters independently name the same farm, that's far stronger than any single page). It's all visible and filterable, so you can sort by how well-evidenced a coffee is.
Roasters can claim their profile and correct mistakes, moving their record from what our AI inferred to what they actually know.
6.Show
Everything becomes explorable: by farm, harvest, roaster, or map.
The connected data becomes the Open Coffee Hub discovery site. Each farm has a page showing which roasters source from it. Each harvest shows all the products built from that green coffee. Each roastery shows the farms behind their menu. And the world map lets you trace the entire supply chain geographically, from origin country to your cup.
Data refreshes automatically every few days. When a coffee sells out it doesn't vanish: it moves to the "Archive" view (the availability toggle) so the trace stays on record.

What we don't do.

We don't make up data. If a roastery doesn't publish a farm name, we mark the harvest as country-only or region-only rather than guess.
We don't rate cup quality. We show verifiable provenance, not a flavor score: any roastery can make exceptional coffee regardless of how much of their sourcing we've traced. How we verify →
We don't take money for coverage. Every roastery we index comes from our own discovery pipeline. There is no paid inclusion.

We curate the roasters ourselves: quality over quantity.

Open Coffee Hub isn't a list of every coffee shop on earth. We keep a hand-picked roster of the world's finest specialty roasters (the ones known for origin-focused, traceable coffee) and concentrate on covering each of them properly.

No pay-to-list, no dragnet. And any roaster can put themselves forward.

An AI confirms each one is a genuine specialty roaster.

Before we invest in reading a whole catalogue, we visit the site and ask an AI: "Is this a specialty coffee roastery?" The AI reads the page, looks for signs of origin-focused roasting (single-farm sourcing, harvest names and periods) and gives a yes or no. Only confirmed roasters move forward.

This step keeps junk out of the database. Chains, tea shops, and supermarket cafes are quietly rejected.

We read every product listing and extract the coffee's story.

For each confirmed roastery, we read every coffee listing and an LLM pulls out the facts that matter, to the depth the roaster actually published: the origin (a named farm, producer, washing station, mill or cooperative, or only a region or country), the process, variety, harvest, altitude, the roaster's cupping score, tasting notes, price, availability and images. Every fact is grounded against the roaster's own words: we keep the verbatim source quote, and anything we can't find in their text is marked unverified rather than shown as fact.

No invented data. If a roaster named only a country, the origin stays country-level, and every fact carries a "how sure we are" marker: stated by the roaster · derived from a name/region · unverified.

We connect coffees from different roasters that share the same farm harvest.

This is the core of Open Coffee Hub. When two different roasters both source from the same farm and the same harvest, we link their products to a single harvest record. We compute a "mapping signature" from the farm name, country, varietals, processing method, and harvest period. Products that share a signature are siblings: different roasters, same green coffee.

That's how you can discover that the Ethiopia Guji harvest from Roastery A and the "Yirgacheffe Natural" from Roastery B are actually the same coffee, priced differently and roasted differently.

We show how sure we are, never a made-up grade.

Rather than invent a quality score, we surface honest, checkable signals: a per-field confidence marker (stated by the roaster · derived · unverified) and cross-roaster corroboration (when several roasters independently name the same farm, that's far stronger than any single page). It's all visible and filterable, so you can sort by how well-evidenced a coffee is.

Roasters can claim their profile and correct mistakes, moving their record from what our AI inferred to what they actually know.

Everything becomes explorable: by farm, harvest, roaster, or map.

The connected data becomes the Open Coffee Hub discovery site. Each farm has a page showing which roasters source from it. Each harvest shows all the products built from that green coffee. Each roastery shows the farms behind their menu. And the world map lets you trace the entire supply chain geographically, from origin country to your cup.

Data refreshes automatically every few days. When a coffee sells out it doesn't vanish: it moves to the "Archive" view (the availability toggle) so the trace stays on record.