Free Tool · UCP Test
The UCP Test
Universal Commerce Protocol is how AI shopping agents discover, read, and buy from stores — and on Shopify, the manifest at /.well-known/ucp now ships automatically. The plumbing is the default.
This test checks the plumbing — then everything that decides whether an agent can actually use your store: the depth of the catalog data it gets back, whether your pages are readable before JavaScript runs, whether a bot wall turns it away at the door, and what AI assistants already say about you today.
What the test checks
Data depth · 45 points
What agents get when they ask
Sampled across up to 50 real products from your public product records — the same records UCP's catalog search serves (we cross-verify that live where stores negotiate): description depth, checkable facts (materials, measurements, fit, care, ingredients), named variants and SKUs, structured attributes, quotable shipping and return policies, and Product schema on product pages. It's what an agent has to work with when it answers a shopper — and still the largest pillar, because no platform fills it in for you.
Page readability · 20 points
Can an agent parse the page at all?
Organization and Breadcrumb schema, titles, descriptions, canonicals and Open Graph cards, citable trust signals — and the one that quietly decides everything: whether your product exists in the HTML before JavaScript runs. Most crawling agents never execute JS, so a client-rendered product page reads as blank no matter how good the catalog behind it is.
Agent access & purchase path · 15 points
Does an agent get in, and can it transact?
robots.txt rules for the ten AI agents doing real consumer shopping, plus what happens when a self-identified agent actually loads the store: bot-challenge walls, password pages, geo gates. Then whether the cart surface is machine-callable and whether records carry the price and live availability an agent needs to commit. Read-only — we never add to a cart or start a checkout.
Protocol · 20 points
Is the manifest there, valid, and live?
Discovery at /.well-known/ucp within an agent's latency budget, strict version-aware validation against the ucp.dev spec, declared checkout/cart/catalog capabilities and payment handlers, and a live probe of the declared MCP endpoint — because a manifest pointing at a dead endpoint reads as a broken store. Worth the fewest points on purpose: on Shopify this layer ships by default.
AI visibility · unscored
What AI actually says about you today
We run shopper-intent queries — your brand by name, your category without it, a real pre-purchase question — through Claude with live web search, and report whether you surface, who gets named instead, and which sources the answer leans on. Deliberately left out of the score: visibility tracks brand fame as much as anything you can change this quarter, and scoring it would make the number unfair to smaller brands.
What is UCP?
The Universal Commerce Protocol is an open standard for agentic commerce, announced in January 2026 and developed by Google together with Shopify, Etsy, Target, Walmart, and a broad group of commerce and payments partners. A store publishes one JSON manifest declaring its protocol version, services, capabilities — checkout, cart, catalog search, fulfillment, discounts — and payment handlers. Any agent that speaks the protocol can then discover what the store supports and transact with it, no custom integration per surface.
It matters because the buyer side is already automated: shoppers ask ChatGPT, Gemini, and Claude what to buy, and those agents complete more of the journey every month. UCP is the sell side's answer — the machine-readable storefront. See the glossary for Universal Commerce Protocol, buyer agent, and brand agent.
The manifest is table stakes. What's behind it is the game.
To be clear: UCP is the best thing to happen to brand-side agentic commerce — an open, shared pipe layer is exactly what lets your data quality matter at all. And here's what most UCP checkers won't tell you: on Shopify, that pipe layer is published for you. Valid version, live MCP endpoint, checkout and cart capabilities, payment handlers — automatic. Which means a passing grade is quickly becoming the baseline, not an edge.
The edge is what an agent finds once it's in the door. Whether your variants are readable or a guess. Whether your differences read as facts or marketing. Whether the answer to “does this ship to me and can I return it?” is right. That's the agent-ready storefront problem, and it's what your brand agent exists to win.
UCP, answered
What is UCP (Universal Commerce Protocol)?
UCP is an open standard, announced in January 2026 and developed by Google with Shopify and other commerce partners, that turns a storefront into a machine-readable surface for AI shopping agents. A store publishes a JSON manifest at /.well-known/ucp declaring its protocol version, services, capabilities (like checkout, cart, and catalog search), and payment handlers. Agents — Google AI Mode and Gemini, ChatGPT, Claude, and platform agents — read that manifest to discover what the store supports and transact with it directly.
What does the UCP Test check?
It walks the path an AI shopping agent walks, and scores each thing that can stop it. Data depth (45 points): the depth of your public product records — the same records UCP's catalog search serves — measured as description depth, checkable facts, named variants and SKUs, structured attributes, quotable policies, and Product schema, sampled across up to 50 products. Page readability (20 points): brand and breadcrumb schema, titles and descriptions and Open Graph cards, citable trust signals, and whether your product exists in the HTML before JavaScript runs. Protocol (20 points): manifest discovery at /.well-known/ucp, spec conformance per ucp.dev, declared capabilities, and a live probe of the declared MCP endpoint. Agent access and purchase path (15 points): robots.txt rules for the agents doing real shopping, whether a self-identified agent gets past bot walls and password or geo gates, whether the cart surface is machine-callable, and whether records carry price and live availability. Plus an unscored AI visibility read: what AI-assisted search says about your brand right now.
Why is data still the largest pillar?
Because the plumbing can't differentiate anyone anymore. Shopify provisions manifests and live endpoints automatically, so competing brands mostly tie on protocol — when we scored 110 well-known brands, the protocol scores of measured stores clustered within a few points. What differs is what an agent gets back when it asks: one catalog answers "will this fit?" and "what's it made of?" with stated facts, the other offers a 30-word vibe. UCP is the pipes; the data is the water. That's why data carries 45 of the 100 points and the protocol layer carries 20 — the weighting is our reading of that reality, and the data checks measure what agents can quote, not a guarantee of how any one agent ranks you.
Do you actually run a browser or complete a checkout?
No, and we're specific about that because it changes what the result means. The purchase-path stage is read-only: it fetches your storefront as a self-identified AI agent to see whether a bot wall, password page, or geo gate turns it away, reads your cart surface to confirm it answers machine callers, and reads price and availability off your public records. It never adds to a cart, creates a checkout, or places an order — so it proves the path is reachable and machine-readable, not that a purchase completes end to end. Everything else in the test is GET requests against public endpoints, and every redirect hop is re-validated as a public host.
How does the AI visibility check work, and why isn't it scored?
We run four to six shopper-intent queries — your brand by name, your category without your name, a real pre-purchase question — through Claude with live web search, then report whether you surfaced, which brands got named instead, which sources the answers leaned on, and what the assistant couldn't find out about you. It is one real AI assistant doing one real retrieval pass; it is not a scrape of Google AI Mode, ChatGPT, or Perplexity, and like real AI answers it varies between runs. That variance, plus the fact that visibility tracks brand fame as much as merchant effort, is why it stays out of the score: a deterministic number you can act on, and honest evidence beside it.
Does the UCP Test actually speak UCP, or just read the manifest?
It speaks it. Kinect publishes its own UCP agent profile at trykinect.ai/.well-known/ucp, and the test presents that profile when it probes a store's declared MCP endpoint — the same capability-negotiation handshake real shopping agents use, as documented in the UCP spec and Shopify's agent docs. When a store completes negotiation, your result shows the live handshake and the exact capability set the store granted. Everything stays read-only: the test runs a sample catalog search and never creates carts, checkouts, or orders — our agent profile deliberately declares catalog capabilities only, so a Kinect session can't transact even in principle.
My store is on Shopify — do I need to set up UCP myself?
Mostly no. Shopify publishes a UCP manifest for stores automatically, with MCP and embedded transports, checkout and cart capabilities, and payment handlers. Run the test to confirm yours is live and complete. The part no platform automates is what agents find once they're in: whether your product data, variants, and policies produce right answers when an agent shops on a customer's behalf.
My grade says "Standard plumbing, thin everything else." What does that mean?
It means your protocol layer is what Shopify ships to everyone — valid manifest, live endpoints, checkout enabled — and the layers above it are thin on the dimensions agents need: short descriptions, few stated facts, policies agents can't quote, pages that need JavaScript before a product appears. It was the most common result when we scored 110 well-known brands, which is exactly the point: at parity plumbing, the brands agents can describe accurately and confidently are the ones whose data, pages, and access answer a shopper's actual questions. The checks under each pillar show precisely where yours falls short.
How is UCP different from ACP (Agentic Commerce Protocol)?
They're parallel efforts. ACP was announced by OpenAI and Stripe in September 2025 and powers checkout inside ChatGPT. UCP followed in January 2026, led by Google with Shopify, Etsy, Target, Walmart, and a broad partner group, and is designed as an open, transport-flexible standard (REST, MCP, A2A) compatible with the Agent Payments Protocol. Most brands will end up speaking both — on Shopify, the platform increasingly handles that for you.
Does Kinect implement UCP for my store?
Kinect's work starts where the manifest ends. The protocol layer is increasingly table stakes — platforms publish it for you, and this free test tells you where you stand. What Kinect builds is the layer agents draw on when they represent you: one intelligence per brand that makes your catalog, what sets you apart, and your policies readable to buyer agents, and an AI sales rep that converts the shoppers on your own site. If the test surfaced gaps you want handled, book a demo and we'll walk your store together.
Buyable is step one. Recommended is the business.
Kinect builds the intelligence behind the protocol — so the agents shopping for your customers find right answers, not guesses. Same-day setup on your live store.