ShopGym: Realistic, reproducible sandboxes for shopping agents

A shopping agent might help you find a jacket within your budget, choose the right size, check the return policy, and add it to your cart. That part feels easy, but training and evaluating these agents is challenging because products, prices, and store layouts keep changing. Bot detection and throttling can also interrupt runs, while hand-built test stores offer control but capture only a limited range of shopping experiences.

This led us to ask: Can we make shopping environments repeatable without losing the complexity of real stores?

We built ShopGym to answer that question. It turns live storefronts into self-contained, resettable sandbox shops and generates shopping tasks tied to their products, features, and policies. Developers can reproduce failures, compare agents under consistent conditions, and measure whether changes improve performance.

This also benefits merchants. Our goal is to build shopping agents that work reliably across stores of all sizes and categories, from clothing and digital goods to home improvement. These environments give agents a place to practice finding products, understanding store policies, and completing shopping tasks before interacting with live stores.