Voice Shoppersimulation
Talk to the marketplace the way you would talk to a voice assistant. State a goal, a budget, a product type and a minimum rating — the shopper plans the request, calls AgentMart's live search_listings, get_listing and get_reviews tools, and shows you every call it made. Nothing here is canned: if the catalogue has no match, the shopper says so.
Conversation
API: checking…Understands: new searches, “cheaper”, “digital only”, “compare 1 and 2”, “shortlist 1”, “details 2”, “reviews 1”. State persists between turns.
Live results
No search yetResults from AgentMart's live catalogue will appear here, numbered so you can refer to them in conversation (“compare 1 and 2”).
Shortlist & comparison
Shortlist at least two listings and the shopper will compare them using live listing details and reviews — highest rating first, then most reviews, then lowest price. The rule is fixed and shown, never hidden.
Tool-call trace
proof, not promisesEvery AgentMart API call this session made, in order. These are the same operations exposed as MCP tools (search_listings, get_listing, get_reviews, list_categories) on AgentMart's MCP server, called here over the equivalent REST endpoints.
AgentMart Voice Shopper — built for the Amazon Developer Hackathon 2026 (Alexa+ track, simulated-experience route). The marketplace launched publicly on October 6, 2026; this conversational layer is the in-window addition. AgentMart is in public beta and settles v1 orders in sandbox credits. Source: web/voice-shopper/ in the AgentMart platform repository.