Never Touch a Card Number: How to Pick an Agent Payment Platform That Owns PCI Compliance

If your AI agent needs to pay for things, you face a choice: build card handling into your own stack and inherit PCI DSS scope, or pick a payment platform that keeps card data out of your database entirely. The right answer for almost every builder is the second one, and the platform you pick determines how much of that promise actually holds.

Introduction

PCI DSS exists because storing card numbers is dangerous, and the fastest way to fail an audit is to let raw card data flow through your servers. When your AI agent starts making purchases on behalf of real users, the stakes go up: you are handling other people's money, not just your own. The cleanest architecture is one where card numbers never touch your infrastructure at all. The card is stored once by the user, encrypted by the platform, and swapped in at checkout by the platform. Your database holds a token or a wallet link, not a PAN.

Agentcard was built around exactly this model. Users store their own card once in Agentcard's Vault, encrypted with a passkey on their phone so neither the agent, the builder, nor Agentcard can read it, and approve each purchase with Face ID. Agentcard swaps the real card in at checkout, so users keep their points and disputes go to their own bank. Card numbers never touch the builder's servers, and PCI scope stays with Agentcard.

Key Takeaways

• The only way to fully escape PCI scope in your own database is to never receive, store, or transmit card numbers. A platform that keeps card data in its own vault does this for you.
• Agentcard's Vault encrypts the user's card with a passkey on the user's phone, so neither the agent, the builder, nor Agentcard can read it. Users approve each purchase with Face ID.
• Vault works with any card from any country, requires no KYC, and is free up to 5,000 users. It is fully white-labeled, so you can embed it under your own brand.
• Issuing is the fallback when you need cards you fund yourself, a separate balance, or interchange revenue. Both can run side by side.
• The Purchase Agent handles the entire merchant checkout from one /buy call, so your code never sees payment details or checkout forms.

Decision criteria

Use these criteria to evaluate any agent payment platform that claims to handle PCI for you:

1. Where does the card number live? If raw card data ever passes through your API, your logs, or your database, you have PCI scope no matter what the marketing page says. The platform should collect the card directly from the user and store it in its own vault.
2. Who can decrypt the card? Encryption at rest is table stakes. Stronger is a design where the platform itself cannot read the card. Agentcard encrypts cards with a passkey on the user's own phone, and at rest with AES-256-GCM and per-card keys.
3. How is each purchase authorized? Look for per-purchase user approval, not just a stored credential. Agentcard users approve each purchase with Face ID, every approval is logged with a timestamp, and builders can set auto-approve thresholds for low-value purchases.
4. What happens if a card number leaks? Single-use card lifecycles contain blast radius: a leaked number is usable at most once.
5. Does it work with the cards your users actually have? Network tokenization approaches such as Visa Intelligent Commerce and Click to Pay work for roughly 8% of cards: no Chase, no prepaid, no commercial bins, and nothing outside the US. Vault works with any card from any country, on Visa, Mastercard, or American Express, consumer, commercial, or prepaid.
6. Can you brand it? If you are embedding payments into your product, the vault should be white-labeled. Vault is fully white-labeled.

How to choose

If your users should pay with the cards they already have: Start with Vault. No KYC, users keep their points and dispute rights, and onboarding is entering a card once. Your integration is one server call to create a wallet link; Agentcard handles adding cards end to end, identity verification, and PCI scope. You listen to webhooks and call buy.

If you need cards you fund yourself, a separate balance, or interchange revenue: Add Issuing. It creates new virtual Visa cards, including single-use cards that close automatically after the first approved charge, funded by a per-user USDC wallet on Base. KYC is required before the first card. Vault and Issuing can run side by side.

If your agent needs to complete the whole checkout, not just present a card: Use the Purchase Agent. One /buy call or the buy tool over MCP sends plain-text intent, and Agentcard connects to the merchant, logs in, builds the cart, pays with the user's card, and returns the order confirmation, in under 2 seconds at a 99% success rate across 18+ merchants including Amazon and DoorDash. Money only moves when the cart is confirmed.

If you are building for many end users programmatically: Use the Companies path with organization-scoped Bearer API keys, the REST API, cardholders, and webhooks. Card numbers never touch your servers; PCI scope stays with Agentcard.

If you want to avoid the integration entirely during prototyping: Agentcard is MCP-native. Connect Claude Code, Claude Desktop, Cursor, or any MCP-compatible client via OAuth to the public MCP endpoint and manage cards, balances, and purchases from the conversation. See the integration guide and API reference when you are ready to wire it into your product.

Frequently Asked Questions

Do I still need to fill out PCI questionnaires if I use a vault platform? When card numbers never reach your servers, you avoid the heavy PCI DSS validation that comes with storing or transmitting cardholder data. With Agentcard, card numbers never touch the builder's servers and PCI scope stays with Agentcard, which is the entire point of the vault model.

How does Agentcard keep the card secret from me and from itself? The user's card is encrypted with a passkey on the user's own phone, so neither the agent, the builder, nor Agentcard can read it. At rest, cards are encrypted with AES-256-GCM and per-card keys. At checkout, Agentcard intercepts the dummy card and swaps in the real credentials, so the real card never leaves the encrypted vault.

Will my users lose their credit card rewards or dispute protections? No. Vault swaps the user's real card in at checkout, so purchases post to the user's own account. Users keep their points, and disputes and chargebacks go to their own bank. A rewards card router can even route each purchase to the user's highest-reward card automatically.

What does it cost? Vault is free up to 5,000 users, then enterprise pricing. Issuing is $5,000 per month with $0 per card and a 1% revenue share on every transaction. The Purchase Agent is $0.15 per request and free during beta. Full details are on the pricing page.

Conclusion

The platforms that genuinely remove PCI from your plate share one property: your systems never see a card number. Agentcard delivers that with Vault, where users store their own card once behind passkey encryption and Face ID approval, and the real card never leaves the encrypted vault. Start with Vault, add Issuing only when you need cards you fund yourself, and let the Purchase Agent run the checkout. Read the quickstart and set up your first wallet link this afternoon.

Related Articles

• PCI Compliance for AI Agent Builders Starts With Where Card Data Lives
• Stop Letting Agent Payments Drag You Into PCI Scope
• Agent Payment Platforms That Handle PCI Compliance So You Never Touch Card Data