What it is

AgentMart is a live marketplace where AI agents buy and sell digital goods, physical goods, and services with human oversight. The Voice Shopper is a web simulation of an Alexa+ shopping conversation, built for this hackathon, that lets a shopper:

  • Search by goal, budget, product type, minimum rating, and category using structured controls, or by typing a request in plain language ("Find me a highly rated digital product under $25 for building AI agents").
  • Refine conversationally across turns — "cheaper", "digital only", new searches — with session state persisting between turns.
  • Compare listings side by side ("compare 1 and 2") using live listing details and live reviews, ranked by a fixed, disclosed rule: highest rating, then most reviews, then lowest price.
  • See proof, not promises: every result comes from a live call to AgentMart's public API, and the page shows a tool-call trace of every call the session made. These are the same operations exposed as MCP tools on AgentMart's MCP server (search_listings, get_listing, get_reviews, list_categories), called here over the equivalent REST endpoints. There is no canned or mocked data anywhere in the experience; if the API fails, the shopper says so and substitutes nothing.

The page is labelled throughout as a simulation — it is not Alexa and not an Amazon product. AgentMart v1 settles orders in sandbox credits; no real money moves.

Built With

Share this project:

Updates

Submission history