NUR.

NUR is a multi-agent shopping assistant embedded in 7TECH's production mobile app — an e-commerce platform for home appliances. Visitors talk to it in any of four languages; it searches the catalog, consults on choices, tracks orders, and completes checkout without the user ever leaving the chat.

The conversation is not text-only. NUR answers with interactive widgets — product cards, comparisons, order status, a working checkout — so the chat behaves like a storefront, not a FAQ bot.

The first design was the obvious one: a single capable agent with many tools. In production it was too slow and hallucinated too often — one prompt carrying every responsibility degrades all of them.

The redesign put a triage-and-routing layer in front of 18 narrow agents: triage classifies intent and hands the query to exactly one specialist — product search, orders, consultation, and others — each with a small prompt and only the tools its job needs. Latency dropped and hallucination frequency fell, because every agent reasons over a fraction of the original context.

NUR ships inside the Flutter app through a WebView + JS bridge, so the agent UI iterates at web speed while the host app stays native. The same bridge pattern carries cart state and auth between the chat and the app.