Virtual cards
Virtual Cards API: Issue Scoped Virtual Cards for AI Agents
Hand an agent a card, not your card. Agentspay issues scoped virtual cards that are merchant-locked, single-use, and time-bound, so a leaked number is worthless beyond its scope.
Pick an agent
Payment intent
intent: ▌
Policy evaluation
Human approval required
This spend is over your approval threshold. Approve it to issue a scoped card, or deny it.
Scoped virtual card issued
Wallet budget
spent of
Audit trail
In short
Virtual cards for AI agents are programmatically issued card credentials that are scoped to a single merchant, amount, and time window. The agent uses a tokenized, masked reference and never sees a raw card number, which keeps you out of PCI scope. You issue and revoke them through a virtual cards API.
How it works
What Virtual cards gives you
Merchant-locked and single-use
A card works only at the merchant and for the amount you allowed, then expires. A loose card cannot be reused or moved elsewhere.
No raw card data in the agent
Agents only ever handle a masked, tokenized reference, so a compromised agent leaks nothing useful and you stay out of PCI scope.
Issue and revoke by API
Create a scoped card in a few calls and revoke it instantly. The card issuing API is sandbox-first so you can test the whole flow.
It is one surface in the flow. Virtual cards works alongside the other control surfaces so every agent transaction is checked against policy before any money moves. See the full picture on how it works and the security page.
Also covers virtual card issuing api, virtual credit card api, virtual cards for ai agents, issue virtual cards to agents.
Further reading: AI agent governance beyond a single card, how to give an agent a virtual card, single-use versus delegated cards, Stripe Issuing for agents explained, and how Stripe agentic commerce scopes agent payments.
Frequently asked
Questions people ask about Virtual cards
How do you issue a virtual card to an AI agent?
Call the virtual cards API with the scope you want: merchant, amount ceiling, and expiry window. You get back a tokenized reference the agent uses at checkout. The raw number never enters the agent context, and you can revoke the card with a single call the moment the task finishes.
Are virtual cards safe for AI agents to use?
A scoped virtual card is far safer than a shared corporate card, because a leaked number is worthless outside its merchant, amount, and time window. What it does not do on its own is decide whether the purchase should happen. Pair it with policy checks and an approval step for that.
What is the difference between a virtual card and a payment token?
A virtual card is a real card credential with its own number, usable at any merchant that takes cards. A payment token, such as the delegated token in an agentic checkout protocol, is bound to one merchant transaction and only works where that protocol is implemented. Cards reach more of the US merchant base.
Does using a virtual cards API put us in PCI scope?
Not if the agent only ever handles a masked, tokenized reference and the raw PAN stays with the issuer. That is the design here. The scope question turns on whether card data touches your systems, so keep the number out of agent context, logs, and prompt history and it stays with the issuing platform.
What are scoped virtual cards?
A scoped virtual card is a card credential issued with its permissions fixed at creation: one named merchant, one amount ceiling, and one expiry window. Outside that scope the number simply declines. The scope is enforced at authorization by the issuer, not by the agent choosing to behave, which is what makes it a control rather than a guideline.
How many virtual cards should each AI agent get?
One per task, not one per agent. A card that lives as long as the agent accumulates risk with every job it runs, while a card minted for a single purchase and revoked on completion has a blast radius of exactly that purchase. Issuing is an API call, so per-task cards cost you nothing but the call.
Can virtual cards enforce a monthly budget across agents?
Not on their own. Each card caps one transaction or one merchant relationship, and no card knows what its siblings already spent. Twenty individually valid cards still total a number nobody approved. The cumulative budget has to live in the wallet layer above the cards, which is where Agentspay enforces it.
Keep exploring
More control surfaces
AI agent wallet
Agent wallet
A wallet is the unit of trust for agent money. Give each agent its own funded wallet with a hard budget, and it can only ever spend what is inside, never your full balance.
ExploreSpend controls
Spend controls
Spend controls are what stop a runaway agent. Set caps, allowlists, and velocity rules once, and Agentspay enforces them on every transaction before any money moves.
ExploreHuman approvals
Human approvals
For the spend that matters, you want a person in the loop. Set a threshold and any transaction above it is held for a one-tap approve or deny before it settles.
ExploreAudit trail
Audit trail
When finance asks which agent spent this, you need one answer. Agentspay writes every transaction to an immutable audit trail, traced to intent, agent, owner, and policy.
ExplorePut Virtual cards to work
Wire it up in the sandbox today, with the SDKs you already use. Agentspay never moves money without policy.
Never moves money without policy