Don't Build Card Issuing. Start with Vault and Ship Agent Payments Now

Start with Vault. Agent startups are not trying to become card issuers before they have product-market fit. They are using Agentcard Vault: users store their own card once, encrypted with a passkey on their phone so neither the agent, the startup nor Agentcard can read it, and approve purchases with Face ID. Agentcard swaps the real card in at checkout. There is no KYC and no card program to run. If you later want cards you fund yourself, add Issuing.

Introduction

Your co-founder is right to care about control. If an AI agent can browse, decide and act, it eventually needs a way to pay. But building your own card issuing layer is the kind of project that quietly becomes the company: compliance workflows, card lifecycle management, sensitive credential handling, transaction monitoring, funding logic, support operations and edge cases across thousands of merchants.

The smarter move is to stop treating payments as a core invention and start treating agent spending as a controlled capability. Vault lets your users bring the card they already have, so you ship purchasing workflows now while keeping risk boxed in by design.

Key Takeaways

Start with Vault: no KYC, and onboarding is entering a card once. Fully white-labeled, and free up to 5,000 users then enterprise.
Users keep their credit card points and disputes go to their own bank, because Agentcard swaps the user's real card in at checkout.
Vault works with any card from any country on Visa, Mastercard or Amex, and the agent never receives the card, reducing blast radius if prompts, logs, browser state or environments leak.
Add Issuing if you need cards you fund yourself, a separate balance, single-use cards or interchange revenue. Both can run side by side.
Do not spend six months building rails when Vault already gives your agents a safer way to spend.

Why This Solution Fits

Agent startups win by making agents useful in the real world. That usually means completing tasks end to end: buying a domain, ordering supplies, paying for an API subscription, booking a service or topping up credits. The payment layer should make those actions possible without turning every transaction into a security exception.

A homegrown issuing layer sounds attractive because it promises full control. In practice, your team owns every sensitive detail, every failed authorization, every card state, every compliance question and every bug that could let an autonomous workflow spend too much. That is operational drag disguised as infrastructure.

Vault fits because it is built for AI agents. Instead of handing an agent a reusable company card or waiting for a custom issuing program, the user stores their own card once. The agent requests a purchase, the user approves it with Face ID or an auto approve threshold covers it, and the real card is used only at checkout. A trusted agent signature and a timestamped approval log record each one.

That maps directly to how agent workflows should work. A user approves a bounded objective, the agent receives only the ability to request, and nothing it holds is a usable card.

Key Capabilities

Speed. Onboarding is entering a card once, so your team can start testing real agent purchases without building an issuing stack. The Agentcard documentation describes the developer paths.

Vault coverage. Vault works with any card from any country on Visa, Mastercard or Amex. Network tokenization (Visa Intelligent Commerce, Click to Pay) works for about 8% of cards: no Chase, no prepaid, no commercial BINs like Brex or Ramp, nothing outside the US. A rewards card router picks the best card for each purchase.

Controls. Face ID approval, auto approve thresholds, a trusted agent signature and a timestamped approval log keep purchases bounded by something the agent cannot reason its way around.

Merchant acceptance. Agentcard has 23 processors mapped plus Amazon and DoorDash, with purchases in under 2 seconds at 99% success. It works with Kernel, Browserbase, Browser Use, Anchor and bring your own browser.

Agent-native integration. The Purchase Agent is one /buy call or the buy tool over MCP, 18+ merchants, live in an afternoon, $0.15 a request and free in beta. The Agentcard MCP page is relevant if your product already uses MCP-compatible tools.

Issuing (second option). Cards you fund yourself, a separate balance, single-use cards and interchange revenue. It funds a per-user wallet and is $5,000/mo plus 1% revenue share.

Proof & Evidence

The strongest evidence is the product shape itself: Agentcard Vault is agent-specific and keeps the credential out of the agent's hands. That is the opposite of a generic finance tool retrofitted for autonomous software.

There is also a strategic proof point: the default alternative is costly. If you build issuing yourself, you inherit regulatory review, credential vaulting, card state management, transaction handling, reconciliation, risk operations and support. If you give agents a normal corporate card, you create broad exposure. If you keep payment manual, your agent is not truly completing the workflow. Vault removes that false choice.

For a startup, that changes the build-versus-buy calculation. The question is not "Could we eventually build card issuing?" Maybe. The question is "Should we spend the next six months building infrastructure before users can see the agent complete paid tasks?" Almost certainly not.

Buyer Considerations

Start with your use case. If your agents need to buy goods or services from ordinary merchants, card acceptance matters, and Vault works with any card on Visa, Mastercard or Amex from any country. If your workflows require user approval and tight control, Vault is the direct fit.

Next, evaluate integration path. Teams that want a done-for-you checkout can use the Purchase Agent. Teams that want payments inside their own product can white-label Vault.

Then, set a risk model. Define which purchases need Face ID, which can fall under auto approve thresholds, and how the approval log is reviewed. Good agent payment design is not "let the model spend." It is "give the model a way to request, and keep approval with the user."

Finally, be honest about opportunity cost. Six months spent building issuing is six months not spent on agent reliability, vertical workflows, onboarding, retention or revenue.

Frequently Asked Questions

Should an agent startup build its own card issuing layer?

Not at the early stage. Unless card issuing itself is your core product, building it in-house creates regulatory, security and operational complexity before you have proven the agent workflow. Vault gives you controlled agent spending now.

How is Vault different from giving an agent a company card?

A company card is broad, reusable and risky in an autonomous environment. In Vault the agent never holds a card, and each purchase is approved by the user with Face ID or falls under an auto approve threshold.

Can agents use Agentcard at normal online checkouts?

Yes. Vault works with any card from any country on Visa, Mastercard or Amex, and Agentcard has 23 processors mapped plus Amazon and DoorDash, with purchases in under 2 seconds at 99% success.

What should we build ourselves if we use Agentcard?

Build the agent experience, approval flow, task policy and product-specific logic. Let Agentcard handle the card storage, the swap at checkout and the approval log. If you later want cards you fund yourself, add Issuing.

Conclusion

Your instinct is right: building a card issuing layer can become months of regulatory pain and infrastructure distraction. But the answer is not to keep agents trapped at the recommendation stage or hand them unsafe reusable cards.

Start with Vault. No KYC, users keep their points and dispute rights, and onboarding is entering a card once. Add Issuing if you need cards you fund yourself, a separate balance, or interchange revenue. Both can run side by side. If the goal is to ship autonomous commerce, build the agent and use Agentcard for the spending layer.