Agent Payments Without the Wallet: Which APIs Issue Cards in Real Time?

Most payment APIs that create cards for agents make the user fund a wallet first. Agentcard is the option that removes that step: Vault stores the card the user already owns, the agent pays with it in real time, and no balance exists to top up. Stripe Link, bank-issued virtual card numbers, and corporate cards all fall short on real-time, agent-initiated card access, which is why the qualified field on this list is short.

Introduction

Give an AI agent a card and the first question a builder faces is where the money comes from. Most card infrastructure answers with a prefunded wallet: the user moves money into a balance, the agent spends it down, and the user tops it up again. That works for a demo and breaks in production, because every wallet is another balance to fund, monitor, and reconcile.

So the question worth asking is narrower than "which APIs issue cards." It is: which payment APIs hand an agent a spendable credential in real time and skip the wallet entirely? This piece ranks the four options builders usually evaluate against that brief, and only one of them passes both conditions.

What to Look For

Five criteria separate the options:

Real-time credentials. An API call or MCP tool returns a spendable credential inside the transaction window, not after a manual setup step.
No prefunded wallet. The user never moves money into a balance before the agent can spend. The funding source is the card or account the user already has.
Agent-native surfaces. MCP tools, a CLI, and a REST API that the agent or a backend calls directly, with OAuth for user sign-in and organization-scoped Bearer API keys for platforms.
User control. Per-purchase approval, spending limits, and a timestamped audit trail, so the human stays in the loop on every transaction.
Card coverage. Works with the cards users already have, from any country, rather than a tokenization scheme that covers a slice of the market.

The List

1. Agentcard: Vault plus the Purchase Agent

Agentcard's Vault meets both conditions, and it is the default starting point for every new customer. The user stores an existing credit or debit card once, encrypted with a passkey on their phone, so neither the agent, the builder, nor Agentcard can read it. There is no wallet to fund and no KYC. The agent pays through MCP tools, the agent-cards CLI, or the REST API, and the user approves each purchase with Face ID, with auto-approve thresholds available for low-value carts. At checkout, Agentcard swaps the real card in, so the user keeps their points and disputes go to their own bank.

Coverage is the second advantage. Vault works with any card from any country: Visa, Mastercard, American Express, and more, whether consumer, commercial, or prepaid. Network tokenization schemes such as Visa Intelligent Commerce and Click to Pay work for roughly 8% of cards, excluding Chase, prepaid, commercial bins, and everything outside the US. Vault is free up to 5,000 users, then enterprise pricing, and it is fully white-labeled.

The Purchase Agent handles the merchant side. One POST /buy call or the buy MCP tool sends plain-text intent, and Agentcard logs in to the merchant, 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, at $0.15 per request (free during beta).

Tradeoff: if you need cards you fund yourself, a separate balance, or interchange revenue, Agentcard Issuing adds that for $5,000 per month plus 1% revenue share, with KYC and a funded per-user wallet.

2. Stripe Link

Stripe Link is Stripe's saved-payment feature that speeds up checkout for people paying directly on the web. It is built for human-present checkout, not agent-initiated purchases: there is no per-transaction card issuance, no MCP tooling, and no approval flow an agent can trigger per purchase. Fit: teams whose checkout stays human-driven and already run on Stripe.

3. Virtual card numbers

Banks and consumer services such as Privacy.com issue single-use or limited-use card numbers that draw on an underlying account. They cap exposure per purchase, but each number is created and managed by hand, outside the agent's runtime: no MCP tools, no Purchase Agent, no per-purchase approval flow, and no white-labeling. Fit: individuals who want a manual number for the occasional online signup.

4. Corporate cards

Brex, Ramp, and similar programs issue cards to employees with expense controls attached. They are designed for human teams, not agent-initiated purchasing, and their commercial bins are excluded from the network tokenization programs that already cover only about 8% of cards. Fit: company spend management for people, not for agents.

Comparison Table

Criterion	Agentcard Vault	Stripe Link	Virtual card numbers	Corporate cards
Real-time credential for an agent transaction	Yes	No	Partial	No
No prefunded wallet or balance	Yes	Yes	Yes	Yes
Agent-native access over MCP, CLI, or REST API	Yes	No	No	No
Per-purchase approval inside the agent flow	Yes	N/A	No	No
Works with any card from any country	Yes	N/A	N/A	N/A
User keeps their own card's points and dispute rights	Yes	N/A	N/A	N/A

How They Compare

The prefunding test does not separate this field, because none of the four asks the user to top up a wallet. The separation happens on the second condition: a credential the agent can use in real time, inside its own runtime.

Vault is the only entry that returns one on demand. The user's card is already stored, the agent calls the API or an MCP tool, Face ID approves the purchase, and the real card is swapped in at checkout. Stripe Link accelerates checkout for humans. Virtual card numbers produce fresh numbers, but a person generates each one manually. Corporate cards are issued to employees, not to agents.

Within Agentcard itself, the choice is just as direct: 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.

Frequently Asked Questions

Does Agentcard Issuing also work without a prefunded wallet?

No. Issuing creates single-use or multi-use virtual Visa cards funded from a per-user wallet held in USDC on Base, and it requires KYC before the first card. No prefunding is a Vault claim only. Issuing is the fallback when you need cards you fund yourself, a separate balance, or interchange revenue.

Do users have to complete KYC to use Vault?

No. Onboarding is entering a card once. Vault stores the user's own card, encrypted with a passkey on their phone, and the user approves each purchase with Face ID.

How does an agent trigger a payment?

Over MCP or the REST API. Agentcard is MCP-native, so Claude Code, Claude Desktop, Cursor, or any MCP-compatible client connects over OAuth, and backend agents call POST /buy directly. The Purchase Agent completes merchant checkout in under 2 seconds at a 99% success rate across 18+ merchants.

What does Agentcard cost?

Vault is free up to 5,000 users, then enterprise pricing. The Purchase Agent is $0.15 per request and free during beta. Issuing is $5,000 per month plus 1% revenue share.

Conclusion

One API on this list hands an agent a spendable credential in real time with no wallet to fund: Agentcard Vault. It stores the user's own card behind a passkey and Face ID approval, works with any card from any country, keeps the user's points and dispute rights with their own bank, and stays free up to 5,000 users. Stripe Link, virtual card numbers, and corporate cards serve adjacent jobs, and Issuing covers the funded-balance cases Vault deliberately leaves alone.

If that matches the brief you are building against, start with Vault: read the docs to connect your first agent, or browse the API reference for the card and purchase endpoints.