GET /.well-known/ucp
Agent-ready commerce
Discover how AI shopping agents see your store, and fix what's blocking your sales before it costs you customers.
UCP profile + catalog · No account required
Free. No card required.
The report
What you get
One overall score, and a point-by-point breakdown of what works and what blocks — each with the fix to apply.
boutique-exemple.com
UCP discoverability
Your store is visible and ready for purchases via AI agents (ChatGPT, Google AI Mode, Gemini…). Keep monitoring this configuration over time.
Catalog quality
Your catalog is usable by a shopping agent: complete, consistent products, ready to be presented with confidence.
Shopping agent simulation
A shopping agent finds your products and fills a cart, but cannot complete the order on its own: a human has to step in, and the sale often goes elsewhere at that point.
Real audit run by the engine — genuine scores, store name withheld.
How it works
Three steps, nothing to install on your store.
- 01
We scan your UCP profile
We query your store the way an agent would: profile present or missing, declared capabilities, catalog actually readable.
- 02
We simulate a real shopping agent
An agent searches your store for a product, adds it to a cart and opens a checkout, to measure how far it can get. No payment is ever triggered.
- 03
You get an actionable score
Every blocker is named, along with what to fix — plus a shareable report for your agency or your developer.
And it stops regressing
A good score is lost without warning, on the first theme or catalog change. Wire the audit into your CI and it becomes a test that fails, not a report nobody reads.
Check failed — score below the requested threshold.--fail-under 85
Illustrative example — simulated numbers, not a real audit.
Badge to paste in your README
It shows the store's latest known score and updates on its own.
[](https://ma-boutique.com)
The GitHub Actions workflow
Exactly as documented in the CLI README — nothing to adapt.
- name: cabfy audit
env:
CABFY_API_KEY: ${{ secrets.CABFY_API_KEY }}
run: npx cabfy audit https://ma-boutique.com --fail-under 85Pricing
The diagnosis is free and unmetered. You only pay for the agent simulation.
Free
€0
To check your store and fix it, as many times as you need.
- UCP discoverability, unlimited
- Catalog quality, unlimited
- Shareable report
- API key, CLI and CI badge included
Growth
€29/month
To measure what an agent can actually buy from you.
- Everything in the Free plan
- 150 shopping-agent simulations per month
- Transaction depth: how far an agent gets into an order
- Cancel any time
The agent simulation runs a real language model against your store: it is the only part of the audit with a variable cost, and therefore the only part that is billed.
Frequently asked questions
What we get asked the most before a first audit.
What is UCP?
Universal Commerce Protocol: the standard an AI shopping agent uses to discover a store, read its catalog and place an order — without ever rendering your site. A store that exposes no UCP profile is simply invisible to those agents.
How is this different from SEO?
SEO optimises what a human sees after clicking a link. Here there is no page and no click: the agent reads structured data and decides on its own what to offer its user. A store with excellent SEO can be completely unusable by an agent — and the reverse is true too.
Is my Shopify store supported? What about WooCommerce or PrestaShop?
The audit works on any store that exposes a UCP profile, whatever the platform. In practice that mostly means Shopify today, where the protocol is the most widely deployed. On WooCommerce or PrestaShop the audit still runs and will tell you plainly that no profile is exposed — native support for those platforms will come later.
How much does an audit cost?
UCP discoverability and catalog quality are free and unmetered, with no card required. Only the shopping-agent simulation is billed, because it genuinely runs a language model against your store — see the pricing table above.
Does the simulation place a real order on my store?
Never. The agent creates a real cart and opens a real checkout to measure how far it can get, then stops dead: no payment is triggered, no card data is handled, and test carts are abandoned. That is a hard rule of the engine, not a configuration option.
How long does an audit take?
A few seconds to a minute for discoverability and catalog. Allow one to two minutes more with the agent simulation, which runs several dialogue turns against your store.
How is the score calculated?
It is the average of the checks that ran, each scored out of 100: above 85 you are fine, between 60 and 84 there are things to fix, below 60 something is blocking. The catalog is judged on a product sample drawn deterministically and replayed identically for as long as your catalog has not moved — so that a score change signals a real change, not a sampling artefact.
Can I wire it into my CI?
Yes. The CLI runs with no install and fails the job when the score drops below the threshold you set; a public badge shows the latest score in your README; and the API is driven by a key generated from your account. All of it is available on the free plan.