ResourcesGuide
Shopify's Claude commerce example: what its storefront agent actually does
Shopify's sample builds a custom storefront agent over UCP catalog, policy, cart and checkout tools. Inspect the code and test one product question before confusing it with Claude channel distribution.
3 min read
Shopify has published Claude for Commerce examples. For a merchant, the key distinction is where this code runs: its storefront sample helps a developer build a custom shopping agent for a Shopify store. It is not evidence that a public Claude consumer shopping service has indexed your catalog or will recommend your products. Shopify’s repository also has a separate merchant agent over the Admin API; this article examines the shopper-facing storefront sample.
The source below is Shopify’s README and its storefront backend inspected on 1 October 2026. It is a code reading, not a live StoreKnows or merchant-store experiment.
Follow one product question through the sample
The backend code calls Shopify UCP’s search_catalog for a query. For a selected item it calls get_product, with lookup_catalog as a fallback. The mapping retains product title, description, price, variants, option labels and an availability flag. That tells you which values this sample tries to read; it does not certify that your product’s unusual compatibility rule is present or correctly interpreted.
The README says the sample also answers from store policies and FAQs. A question such as “Can I return this after 30 days?” belongs to that source rather than a camera lens compatibility table. The code’s policy/FAQ call is distinct from its catalog calls. Review the actual response and source text before deciding whether a failure is a policy, product-data or agent-reasoning problem.
Cart and checkout have a clear handoff
The sample can create, inspect and update a UCP cart. Its backend re-reads the cart before composing an update, because an update replaces the line list and the storefront may have changed the shared cart. For checkout, it stages a UCP checkout and returns a continue_url. The README explicitly says the buyer finishes on Shopify’s hosted page; the sample does not call complete_checkout, place an order or take payment.
Optional Sign in with Shop adds a buyer-linked catalog context. Without the configured app credentials, the README says sessions use the guest path. Optional order questions have their own scope and are limited to orders placed through this agent, rather than a general customer history. These details matter if you promise personalization or order support from your own deployment.
A bounded test for your store
If you are considering the sample, take one important product question and record:
Scroll sideways for columns out of view. Keyboard: focus the table and use the Left and Right arrow keys.
| Record | What to capture |
|---|---|
| Maintained fact | Exact product/variant, source, owner, conditions and update date. |
| Configuration | Store domain, sample revision, guest or signed-in path, market and enabled permissions. |
| Tool response | Catalog or policy response actually returned for the question, with sensitive data removed before sharing. |
| Agent answer | Exact prompt, response, date, and whether the answer carried the source condition. |
| Checkout test | Whether the correct variant reached the cart and Shopify hosted checkout; do not place a real order merely to demonstrate an answer. |
If the source is absent, fix the source or its native Shopify representation first. If the tool response includes it but the answer drops the qualifier, that is a different failure. A successful local run says only what this sample did with this store and configuration. It says nothing about product placement in other AI channels.
If you simply want a managed on-store chat option, the README points to Shopify Inbox. For AI channel distribution, check Shopify Agentic Storefronts and the relevant channel settings. A StoreKnows simulated check may help assess a supported website-tool question; it is not required by this UCP sample and cannot turn the sample into a public Claude distribution channel.
StoreKnows is independently developed and not affiliated with, endorsed by or sponsored by Anthropic or Shopify.
Inspect the sample and its revision
Record the code revision you inspect and follow one product question through the sample before making a claim about its behavior.
Open Shopify's Claude commerce sampleQuestions this article answers
- Does using Shopify's Claude example make my products appear in Claude shopping?
- No such distribution is established by the sample. Shopify presents it as code for a custom storefront agent you run against a store. AI channel distribution and its controls are separate.
- Does the sample complete an order in chat?
- The README and inspected backend say checkout is staged and the shopper follows a Shopify continue_url to finish. The example does not call complete_checkout or take payment itself.