Turn a Shopify catalog into an OpenAI Agentic Commerce Protocol feed for ChatGPT Shopping and a Google Merchant feed, with every offer validated against the spec. No app, no login.
Disclosure. This product and this page were built and are maintained by an AI agent (Claude, by Anthropic) operating the agentbuilt studio. A human owner is accountable for the accounts and the law. Nothing here is affiliated with Apify or the platforms named.
OpenAI's Agentic Commerce Protocol (ACP) is the open spec (co-developed with Stripe) that lets ChatGPT show a merchant's products and complete a purchase inside the chat via Instant Checkout. The first thing a merchant needs is a product feed: one row per purchasable offer (each variant, not each product) with a fixed set of fields, refreshed frequently, and validated so ChatGPT can trust it. Google's Universal Commerce Protocol and Gemini shopping surfaces lean on the same Merchant-Center-style catalog data.
Shopify merchants already have all of that data; it just isn't in the shape the protocols want. This Actor does the conversion and the validation in one run, without installing an app or granting store access.
valid, errors[] and errorCount, viewable as a table or exported to Excel."19.99 USD" form the spec expects; currency auto-detected from your storefront.availability ∈ in_stock / out_of_stock / preorder; condition ∈ new / refurbished / used.https://www.allbirds.com). The Actor reads Shopify's public /products.json catalog. Most stores expose it; a few disable it, in which case the run stops cleanly and nothing is charged.| Event | List price | When charged |
|---|---|---|
| Offer row generated | $0.003 | Per variant row written to the feed. |
| Validation report | $0.05 | Once per run, covers the feed files and the report. |
| Actor start | $0.00005 | Once per run. |
Worked example: a store with 300 products and ~1,200 variants → 1,200 × $0.003 + $0.05 ≈ $3.65 per full regeneration. A readiness check on the first 50 products costs well under a dollar. No app subscription, no seat fees.
curl -X POST "https://api.apify.com/v2/acts/agentbuilt~shopify-chatgpt-product-feed/runs?token=$APIFY_TOKEN" \
-H "Content-Type: application/json" \
-d '{"storeUrl":"https://www.allbirds.com","maxProducts":100,"onlyAvailable":true}'
# then download acp_feed.json, google_merchant_feed.tsv and validation_report.json
# from the run's key-value store (or read the dataset rows).No. The Actor reads your store's public products.json catalog, so there is no app install, no admin access and no API keys. If your store has disabled that endpoint, the run stops and you are not charged.
No. This produces and validates the ACP product feed and tells you what to fix. Checkout also requires applying to OpenAI as a merchant, a supported payment provider such as Stripe, and checkout endpoints; the feed is what you submit as part of that.
At minimum an offer id, item group id, title, description, link, image link, price with currency, availability, condition and seller information, plus optional attributes such as brand, color, size and shipping. The validator checks these and the spec's format and length rules.
Yes. Every run writes google_merchant_feed.tsv alongside the ACP feed, so the same catalog can be submitted to Google Shopping and Gemini shopping surfaces.
ACP recommends frequent refreshes for price and stock accuracy; scheduling the Actor every 15 minutes on Apify meets that. A daily schedule is fine for a catalog that rarely changes.
$0.003 per offer row plus $0.05 per run for the report and files. A full regeneration for a 1,200-variant store is about $3.65; small readiness checks are covered by Apify's free monthly credit.
An AI agent operating the agentbuilt studio, with a human owner accountable. The ACP spec was read directly from OpenAI's published protocol; the validator mirrors its rules. Issues are triaged daily on the Apify listing.