The brief was blunt: the client kept hearing that shopping agents were sending buyers to competitors, and their store was invisible to all of them. The agency had six days of budget and no appetite for a re-platform.
Mid-size Shopify apparel store · 3-person agency team · AgentCart ACP + Bridgehead UCP · 6 days from kickoff to first live agent order.
The problem
Agent surfaces need two things a standard Shopify storefront doesn't expose: a discovery layer that says "this store exists, here is the catalog, here is what's in stock" — and a checkout API an agent can drive end to end. The theme, the apps, the analytics stack: none of it matters to an agent that only speaks UCP and ACP.
Building both protocols from the spec was quoted at six to eight weeks. The agency had six days.
The build
Day one was Bridgehead UCP: the Shopify adapter read the existing catalog and published the discovery endpoints, so agents could find the store and read live inventory by the first evening.
Day two through four was AgentCart ACP: checkout sessions, delegated payment tokens, and the order webhooks wired into the store's existing fulfillment flow. The conformance test suite caught the one bug that would have hurt — a retry path that would have double-created orders — before any agent ever hit it.
// The whole integration surface the agency wrote by hand:
const bridge = bridgehead({ adapter: shopify(store), pairWith: acp });
const acp = agentcart({ catalog: bridge.catalog, fulfill: shopifyOrders });Day five and six were staging runs against real agent clients and go-live.
Results
We quoted the from-scratch build at eight weeks and won the job with six days. The conformance suite is what let us sleep. — Agency lead
The pattern generalizes: Bridgehead answers "can agents find and read the store", AgentCart answers "can they buy from it", and the storefront underneath never knows the difference.
Golam Mostafa builds the ACP, UCP, and MCP products behind these results. Planning something similar? Talk it through first.