Three cultural grocers run on comrse today. Underneath each one: a catalog organized so people and machines can read it, and a salesperson who knows the recipe, the culture, and the shelf.
“Tengo $30 pa' la cena, algo cubano.”
She hears the budget, builds the whole dinner inside it, and answers in the Spanish her customers actually speak: abuela first, algorithm second.
Live, talk to her Visit the market → EN · Pantry“Stock my pantry for the month, keep it under $120.”
A chef who plans the month, respects the number, and knows what actually goes in the same basket, because fifteen years of orders taught him.
Live, talk to him Visit the bodega →“Qué necesito para una paella de verdad?”
A storefront built museum-clean, with a hundred and ninety recipes ready to answer the only question that matters: what do I cook with this?
Opening nowA cultural grocery carries thousands of products a supermarket never will, and the knowledge of what to do with them lives in the owner's head. The website has neither, a search box, some categories somebody made years ago, and silence.
Meanwhile the customer's question moved. It is no longer “do you have sazón” typed into a box. It is “what do I need for lechón for twelve” asked of a chat window, a search engine, or a shopping assistant that will answer with somebody's store. The work below decides whose.
Before anything can find a product, the catalog has to make sense. We rebuild it two-tier: how people shop (the occasion, the dish, the diet) crossed with what things are (category and brand), and retire the dead ends.
On one of these stores that meant collapsing years of accumulated collections into a structure a third the size, with zero broken links left behind.
Structured data on every product page: price, availability, origin, what it is and what it is for, plus recipe pages with real schema, so the catalog earns its way into ordinary search instead of renting its way in.
Including the unglamorous part: hunting down the legacy URLs an old platform left in the index and pointing them home.
The search box is becoming an agent. When a shopping assistant answers “what should I buy,” it recommends catalogs it can parse and trust: machine-readable pages, consistent attributes, honest availability. Most stores are invisible to it.
We build for that reader on purpose: the same taxonomy and schema, plus the files and signals that tell an AI agent it is welcome and what it is looking at.
On top of all of it, a salesperson, not a widget. Recipe knowledge: asked for maqluba or mojo, she knows the ingredient list and pulls it from the shelf. Cultural knowledge: she code-switches, knows the holiday, knows what abuela would actually serve. Catalog knowledge: she sells what is in stock, inside your margins, at the customer's budget.
That is why the same engine shows up as a Cuban aunt on one store and a spice-house chef on another. The fluency is the feature.
Start with the catalog work. It pays for itself in search before the chat ever says hello. Then give the store a voice that sounds like yours.
This is one of six
comrse writes custom software for businesses that sell physical things, in any industry. Every build is made for one company and shaped around how that company makes money, and every one of them starts as a full copy of a platform that has been carrying our own stores for fifteen years. That is why a smaller business can afford work like this without any of it being built thinner.