Żabka Jush Lets AI Agents Build a Shopping Cart From a Description, Recipe or Fridge Photo

Żabka Jush has launched a beta feature that lets users hand off the job of preparing a shopping cart to a compatible AI tool. Users can describe what they need, paste a link to a recipe or send a photo of their fridge.

What the agent can do

The user connects a compatible AI tool to their Żabka Jush account. They can then describe what they need, paste a link to a recipe or upload a photo of their fridge. The AI searches for products, can factor in promotions and coupons, and finally prepares a cart.

The tool can also plan shopping for several days, taking into account the diets, allergies and preferences of household members. For a few items bought in a hurry, the cart is created in Jush. For larger, planned shopping, the AI can assemble it in delio supermarket online.

How it works technically

The whole thing is built on the Model Context Protocol. With the user's consent, AI agents can use selected Jush and delio functions, such as searching for products, checking promotions and preparing a cart. The solution runs as a Remote MCP Server, so according to the source it does not require installing any additional software.

It is part of a trend known as agentic commerce, in which AI does not just answer questions but also carries out the follow-up steps needed to complete a task. The user decides whether to connect an AI tool to their account.

Limits of automation

The feature has clear limits. It is in beta, and the AI does not place or pay for the order. The cart has to be reviewed and finalized in the app, so the customer has the last word.

Michał Tajchert, CTO at Lite e-Commerce, part of Grupa Żabka (the Żabka Group), says the future of e-commerce will rely less on manually searching for products and more on understanding what the user wants to achieve. Technology is meant to simplify the process and take repetitive tasks off the user's hands.

We want technology to simplify this process and take some of the repetitive tasks off the user, while leaving them with full control over the order. - Michał Tajchert, CTO at Lite e-Commerce, part of Grupa Żabka

What it means for customers

For customers, this is a practical example of grocery shopping with AI agents in Poland. Instead of manually searching for the products in a recipe, users can hand the task to an assistant they already use. The source does not say which AI tools are compatible, or when the beta will end.