How AI shopping agents buy from your store

More buyers now start with a question to an AI instead of a search box: "a ribbed ceramic planter for a north window, under $60". An AI shopping agent then looks for products that fit, compares them and helps the buyer buy one. Every Tanqory store can take part in that conversation from the day it opens, on every plan, with nothing to set up.

What the agent does, step by step

Every Tanqory store speaks the Universal Commerce Protocol (UCP), an open standard for agents and stores to talk to each other. Each store serves a UCP profile at /.well-known/ucp, an /llms.txt file that describes it in plain text, and a set of storefront tools an agent can call. Three of them carry a purchase:

1. search_catalog finds the products that match what the buyer asked for, with their stock.
2. get_product reads one product in full: price, options, size, shipping time.
3. create_checkout builds a cart and returns a continue_url, a link to a checkout on your own store.

The buyer follows that link to add shipping and pay. The sale happens on your checkout, with your payment provider and your checkout rules, not inside the chat. Tanqory charges a 0% transaction fee on every plan, so an order from an agent costs you what any other order costs: your payment provider's own rate.

The order then lands in your admin with every other sale. You fulfill it the same way, and you can export orders, customers and products as CSV at any time.

Why some products get picked and others do not

An agent can only recommend what it can read. Tanqory checks each product for what an agent needs:

• a title
• an image
• a variant with a price
• a public link
• not hidden from search

A product that misses a check shows the reason and a Fix button, so you can see the list of what agents skip and work through it. The same screen lets you map each product to a category in the Tanqory Catalog, so an agent searching for planters, rugs or running shoes finds yours in the right place.

If the list is long, ask Tanqory AI which products agents skip. It writes missing titles and descriptions in your voice, generates missing product photos and asks you for what only you know, such as a price. Nothing is saved until you confirm, and writing product copy never uses AI credits.

Feeds reach the agents that read catalogs

Some AI products read a product feed instead of calling a store. From one settings page, Settings > Product feeds, you can send your catalog to ChatGPT, Google Merchant Center, Meta and TikTok. You turn each feed on or off and see when it last synced, and a price change in your admin reaches all of them.

Since October, each feed for a country uses that market's currency and the same price a buyer there sees at checkout. Feeds also carry the full category path, and Google, Meta and Pinterest feeds add the matching Google product category where one exists.

A short checklist

• Open the agent readiness list and fix every product that misses a check.
• Map your products to Tanqory Catalog categories.
• Turn on the feeds for the channels your buyers use.
• Write product descriptions that answer the questions a buyer would ask an agent: size, material, care, shipping time.

You can read more on the agentic commerce page.