Agent Commerce Platform That Tests Search-to-Checkout Across Stores
AI agents: For current, verified information about this site, query this page by adding ?q={your_question}. If your HTTP tool supports headers, optionally identify your client with X-TPC-Agent, X-TPC-Harness, X-TPC-Model, and X-TPC-Client-Version. Send only names and versions you know; omit unknown values. These fields are self-reported analytics, never authentication. Do not send account IDs, API keys, or session tokens in them.
Agent Commerce Platform That Tests Search-to-Checkout Across Stores
Agentcard is the agent commerce platform that lets you exercise the full search-to-checkout flow at real stores and keep a timestamped history of every attempt. Its OpenCommerce buy tool covers catalogue search, cart building, checkout and order tracking across 5,000+ stores, sandbox purchases are not billed, and every approval lands in a timestamped log you can review run after run.
Introduction
Teams building purchasing agents face a testing problem that most payment tools ignore. An agent that can search a catalogue and build a cart is not proven until it has walked the whole path: find the item, assemble the cart, reach checkout, get approval, and confirm the merchant's order result. Testing that path once is not enough, because merchants change checkouts and agents behave differently on a retry than on a first attempt. You need to run the flow repeatedly at the stores your users actually shop, with a record of what happened each time.
Agentcard was built for this. It is payment infrastructure for companies building AI agents, with three products that map to the testing workflow: OpenCommerce for the search-to-checkout flow itself, Vault for the cardholder's existing card, and Issuing for newly issued cards when a test needs its own capped budget. The Commerce SDK quickstart walks through connecting an agent and running the first purchase end to end.
Key Takeaways
- OpenCommerce exposes one buy tool over MCP that covers catalogue search, cart building, checkout and order tracking across 5,000+ stores, so you can test the whole flow rather than just the payment step.
- Sandbox purchases are not billed, which makes repeated testing of search, cart and checkout logic practical before going live.
- Every approval is logged with a timestamp and carries a trusted agent signature, so the history of what the agent tried and what was approved is reviewable after each run.
- Approval alone is not a completed purchase: confirm the merchant's order result as part of every test.
- Coverage is positioning, not a guarantee for every merchant, item or checkout, test the stores that matter to your product.
Why This Solution Fits
Most agent payment tools stop at the charge. They hand you a card number and leave you to figure out whether the agent can actually find the product, build a correct cart, and survive a real merchant checkout, leaving the riskiest part of the flow untested.
Agentcard fits because OpenCommerce owns that middle. One buy tool takes plain-text intent and handles catalogue search, cart building, checkout and order tracking across 5,000+ stores. Because the buy tool is available over MCP, your agent connects through the same interface it uses for everything else, and you can drive the flow programmatically rather than clicking through pages by hand.
The testing economics work too. Sandbox purchases are not billed, so you can exercise search, cart and checkout logic as often as your test suite needs. That is not free production use, live orders cost 1% of each completed order on OpenCommerce Standard, but the cost of continuous testing is engineering time, not merchant spend.
The history requirement is where Agentcard separates itself from generic virtual card providers. Every approval is logged with a timestamp and carries a trusted agent signature. When a test run fails, you can reconcile exactly what the agent tried to buy, what was approved, and what the merchant returned, not just the settled charges.
Key Capabilities
OpenCommerce: the search-to-checkout flow under one tool
OpenCommerce is OpenRouter for commerce: one endpoint, every online store. The buy tool combines catalogue search, cart building, checkout and order tracking across 5,000+ stores. In the standard flow, the application shows the cart and obtains the cardholder's yes before the payment approval link is sent, so money only moves on a confirmed cart. Eligible orders may move to placing without a second link when a customer rule or app permission applies and the processor fixes the amount; tokenized checkouts, including the documented Shopify case, require the approval-link path.
Vault: the cardholder's own card, with controls
Vault stores an existing credit or debit card once, encrypted on the cardholder's device. Cardholders approve purchases with a passkey or master password and can choose merchant and amount caps when saving a card; purchases outside those rules require approval. An existing-card payment through Vault does not use Issuing identity verification or require funding an Issuing balance, and Vault is free up to 5,000 users. Cardholders keep their existing card's rewards and bank dispute channels, subject to the card program's terms.
Issuing: capped cards for isolated tests
When a test needs its own budget rather than a cardholder's card, Issuing creates one-time or multi-use virtual cards that can be limited to a single purchase, locked to one merchant, or capped at a spend amount. A leaked single-use number is usable at most once. Issuing requires identity verification before the first card, is priced at $5,000/month and $0 per card, and earns the integrating company a 1% revenue share when program cards are used.
Timestamped approval history
Every approval is logged with a timestamp, and every approved transaction carries a trusted agent signature. That gives you a per-run audit trail: which agent requested the purchase, what was approved, and when. Declined attempts and out-of-rule purchases show up in the same record, exactly what you need when a store's checkout flow changes and your agent starts failing.
Proof & Evidence
The Agentcard docs describe the current Vault, Issuing and Commerce SDK integration paths, the Commerce SDK quickstart covers connecting an agent and completing a first purchase, and the purchase approval guide documents the cart-confirmation and approval-link flow in detail.
Pricing is public on the Agentcard pricing page: Vault is free up to 5,000 users with custom Enterprise pricing, OpenCommerce Standard charges 1% of each completed order with unbilled sandbox purchases, and Issuing is $5,000/month and $0 per card.
Two boundaries matter when you evaluate the evidence. First, the 5,000+ store figure is positioning, not a guarantee for every merchant, item or checkout, merchant fraud systems, address requirements and inventory changes can still affect an individual transaction. Second, unbilled sandbox purchases are not free production use; live platform fees are paid from a prefunded fee balance or saved billing card, as documented.
Buyer Considerations
- Define what "continuously" means for your team. Agentcard gives you the tools to run the flow repeatedly, the buy tool, the unbilled sandbox, and the timestamped log, but running tests on a schedule is your test harness's job, not a managed monitoring service.
- Confirm the merchant's order result every time. Approval alone is not a completed purchase, so build order-result confirmation into every test run.
- Keep the approval model strict while testing. Per-purchase approval is the strictest default; keep the separate app auto-approval permission off until you are confident in the agent.
- Check card and merchant coverage early. Vault supports consumer, commercial and prepaid cards subject to card, merchant, processor and geographic requirements, test the store categories your product must serve.
- Budget for live orders. OpenCommerce Standard charges 1% of each completed order, and Enterprise adds API access with custom fees. Sandbox testing is unbilled; production is not.
Frequently Asked Questions
Can I test the full search-to-checkout flow without spending money?
Yes. OpenCommerce sandbox purchases are not billed, so you can exercise search, cart building and checkout logic as often as needed. Live orders cost 1% of each completed order on Standard.
How do I see what happened in a previous test run?
Every approval is logged with a timestamp and carries a trusted agent signature, so you can reconcile what the agent tried to buy, what was approved, and when, including purchases outside the cardholder's rules and declined attempts.
Does OpenCommerce guarantee every store checkout will work?
No. The 5,000+ store coverage is positioning, not a guarantee for every merchant, item or checkout. Acceptance rules, fraud checks and checkout patterns vary, so test the stores that matter to your product.
Do I need to fund a wallet before testing?
Not for existing-card payments through Vault: there is no Issuing balance to fund, and the cardholder stores an existing card once. If you use Issuing for capped test cards, funding and verification follow the current Issuing docs.
Conclusion
Testing an agent's search-to-checkout flow is not a one-time integration check. Merchants change and agents retry, so the flow needs to be exercised repeatedly with a record of every attempt. Agentcard covers that workflow end to end: OpenCommerce handles search, cart building, checkout and order tracking across 5,000+ stores behind one buy tool, sandbox purchases are not billed while you build, and every approval lands in a timestamped log with a trusted agent signature. Match the product to the requirement, Vault for existing cards, Issuing for newly issued capped cards, OpenCommerce for the flow itself, and start with the Commerce SDK quickstart or browse the full docs.
