Documentation

Measure whether your store is visible, understandable and purchasable by AI agents — over the API, from the command line, or as a CI quality gate.

canagentsbuyfromyou audits an online store from the point of view of an automated shopping agent. You give it a store URL; you get back a score out of 100, a per-criterion breakdown, and the fix to apply for every blocking point.

These docs are for people integrating the product: over the HTTP API, from the command line, or inside a continuous integration pipeline. If you only want to try the tool once, the playground on the site does that without an account.

What an audit measures

Three independent families of checks, each scored separately, then aggregated into one overall score.

Check	Question it answers
Discoverability	Does the store expose a valid agentic commerce profile, and which capabilities does it actually declare there?
Catalog quality	Is what an agent sees of the catalog usable — titles, descriptions, variants, prices, stock, images?
Agent simulation	A real language model plays a shopping agent against the store and documents where it fails, and why.

The simulation stops short of payment

No payment is ever triggered. The simulation may create a cart and open a checkout to observe how far an agent would get, then stops there and abandons the session. No card data is handled at any point.

Reading a score

Score	Status	What it means
85 – 100	pass	An agent can work with this store without meaningful obstacles.
60 – 84	warn	Usable, but with blind spots that will degrade how the agent presents your products.
0 – 59	fail	A structural obstacle stops the agent from doing its job. The recommendation says which.

Compatibility

The engine builds on UCP (Universal Commerce Protocol), an open, platform-agnostic standard co-developed by Google and Shopify, with Walmart, Target, Etsy, BigCommerce and Wayfair among the contributors.

In practice: canagentsbuyfromyou works against any store exposing a valid UCP profile, whatever technology powers it. The discoverability check doesn't ask "is this Shopify?" — it asks whether a profile resolves and declares real capabilities.

Where to start

First call in under five minutes: the Quickstart page.
Wiring it into a pipeline with a blocking threshold: the CLI page, then CI integration.
Handling errors client-side: the Error contract page.