A 300-Product Catalog in a Few Minutes: How Claude Filled It

I received two simultaneous orders from the Ukrainian funeral industry: an online catalog of monuments for dealers and SEO for a separate website. I created the catalog using my video feed service, shop.vibevox.pro, and the Claude AI chat, connected via MCP protocol. A few text commands transformed an empty store into a ready-made structure with nine categories and 300 products. A similar principle of connecting the AI to the catalog and transferring products via a link is already built into 1001site AI —you can build your own store this way without a dedicated development team.

Objective: A working tool for dealers

The customer described the task directly: “We need an electronic catalog for the work of our dealers and partners. This is not an advertising catalogue, but a working tool for a wide selection of models.” Dealers and partners themselves use it: they select a monument model from the catalog and place an order.

At the same time, a second order arrived: to promote the existing website pamyatniki.ukr, which I didn't develop. This work began with a comprehensive SEO audit to understand growth opportunities and estimate costs. The rest of the case study focuses solely on the catalog; the second project is unrelated.

How the catalog is organized: from ordering to product card

For the dealer catalog, I used my video feed service, shop.vibevox.pro, a separate product, not 1001site. In the overall system diagram, the catalog is part of a larger process: orders are copied to the manager's Telegram, statuses are tracked as "New → In Progress → Completed," and customers are directed to the store from Telegram, WhatsApp, or Instagram links.

Products in the management system are sorted by category, each can be quickly assigned a description and photo, and the category and card itself are created using a neural network, eliminating manual coding. This step—creating categories and cards without manual input—was accomplished in the monument order by the AI chat via MCP.

MCP: The service connects to Claude as a tool

The open Model Context Protocol connects AI chat directly to an external service: the chat invokes service functions via text commands instead of clicking on the admin panel. Shop.vibevox.pro already has this feature in place—I show it in the screenshot of the service's landing page above, in the cover image of this article.

The service page lists what can be managed through Claude without coding: products, categories, media files, store settings, orders, texts, and SEO descriptions. Connection is via a single link with a personal key, and access is limited to a specific store.

What I really wrote to Claude

Setting the task took just a couple of commands in chat. One of the prompts looked like this—with clear steps and no room for guesswork:

Ты подключён к моему магазину VibeVox через MCP (инструменты vibevox_*). Магазин: cnc. Сделай строго по шагам и в конце покажи итог. ШАГ 1. Создай категории - вызови vibevox_add_categories с полем names (ровно эти 9, с enable: true): ["АНГЕЛИ","ХРЕСТИ","СІМЕЙНІ","ДИТЯЧІ","КВІТИ","ВІЙСЬКОВІ", "МУСУЛЬМАНСЬКІ","ЄВРЕЙСЬКІ","ЕЛЕМЕНТИ ПАМ'ЯТНИКА"] ШАГ 2. Вызови vibevox_list_categories. Для каждой категории запомни её name и slug (slug бери именно из ответа API, не выдумывай). ШАГ 3. Вызови vibevox_get_store_settings. Возьми текущий settings.design.taplinkConfig...

The command doesn't mention anything about the interface or buttons—only specific category names and the order in which the service functions are called. These specifics make all the difference between "asking the AI to come up with something" and "asking the AI to perform a precise task": the more precise the steps, the more predictable the outcome.

What Claude answered: 9 categories in 6 steps

Claude completed the task and reported on each step separately: he created 9 categories without duplicates, and rebuilt the store's design settings (background color, layout) because the old links weren't on the site yet—there was nothing to lose.

The response contains a ready-made table: the number, category, technical slug, and a working link to the store section for each of the nine categories (from "Angels" and "Crosses" to "Monument Elements"). Claude generated the slugs himself using transliteration and explicitly stated that he took them from the API response and didn't make them up—this is exactly the kind of control I requested in the prompt.

Claude later used the same approach to sort through the entire array of original 3D models: categorizing them into nine categories and converting them into 300 individual product catalog entries. Dealers received a ready-made, structured tool instead of a solid list of models.

Background videos and project summary

To prevent the catalog's main page from looking like a boring list, I generated background videos for each category. This way, the interface visually distinguishes "Angels" from "Military" monuments even before opening the section. The post itself includes a video demonstration of the interface. At the time of publication, the result was already being approved by the client.

You can check out the final product in the live store - shop.vibevox.pro/cnc. I estimate that about 5% of the project's manual work remains—model analysis, catalog structure, and content were handled by the AI chat.

You can also create a similar catalog with an AI-assisted dashboard on your website using 1001site AI: a style catalog, product transfer via link, or MCP connection to Claude, ChatGPT, or Gemini.

Sources

• Post from the "Business with AI" channel: https://t.me/promtvibeadd/408
• VibeVox video streaming service: https://shop.vibevox.pro/
• Dealer directory result: https://shop.vibevox.pro/cnc