Tools

The Vistoya MCP server exposes five read-only tools. All tools return JSON text content and never modify data.

Product discovery

Tool	Best for
`discover_products`	Natural language queries — AI semantic search with vector embeddings
`find_similar`	"More like this" — find alternatives to a specific product

Product details

Tool	Best for
`get_product`	Full product info — variants, images, AI attributes, purchase link

Catalog metadata

Tool	Best for
`list_stores`	Browse all connected retailers with product counts
`get_filters`	Available categories, brands, colors, materials, styles, silhouettes, price range

Choosing the right tool

Use discover_products for all product search — whether the request is descriptive ("something elegant for a cocktail party"), filter-based ("sneakers under $150"), or anywhere in between. It uses AI semantic search backed by Qdrant vector embeddings and handles both natural language and structured intent well.

Call get_filters first at the start of a session so the agent knows what categories, brands, colors, and materials are available in the catalog.

Response format

Every tool returns a content array with a single text entry containing JSON:

{
  "content": [
    {
      "type": "text",
      "text": "{ ... JSON payload ... }"
    }
  ]
}

On errors, the response includes "isError": true with a descriptive error message.