Stripe Issuing for Agents: Virtual Cards, Pricing, Link Wallet and Spend Controls

Short answer: Stripe Issuing for agents lets you issue single-use virtual cards that AI agents can spend on autonomously, with the credential scoped by amount, currency and merchant category, plus transaction monitoring and fraud tools. On top of it, Stripe launched a Link wallet for agents that gives an agent programmatic access to a one-time-use card or a Shared Payment Token backed by the cards and bank accounts already in a user's Link wallet, with the user approving the spend request first. It is strong issuing and wallet infrastructure. It is not, by itself, a rail-neutral buyer-side control plane that sets per-agent policy, gates high-value spend for human approval, and keeps one audit trail across every rail you use.

What Stripe announced

At Sessions 2026, Stripe rolled out the pieces that let agents pay for things. Two are worth understanding separately because they solve different parts of the problem. Issuing for agents is the card infrastructure: it provisions virtual cards an agent can charge, with spend controls attached to the credential. The Link wallet for agents, announced at the end of April 2026, sits a layer up: it lets an agent tap into a consumer's existing Link wallet and get either a one-time-use card or a Shared Payment Token that draws on cards and bank accounts the user already saved.

Put simply, Issuing for agents is how a business gives its own agents cards to spend on. The Link wallet for agents is how a consumer lets a third-party agent pay with the payment methods already in their wallet. Both are real, both shipped in 2026, and they answer "how does an agent get a usable payment credential," which is a genuine gap the category needed filled.

How the spend controls work

The controls live on the payment credential. When you issue a card for an agent through Issuing for agents, you can scope it by amount, by currency, and by merchant category. Merchant category controls are the useful part in practice: a travel-booking agent can be locked to airlines and hotels and blocked from restaurants or retail, so even if the agent misbehaves it physically cannot charge outside its lane. Stripe adds real-time authorization decisioning, transaction monitoring and its fraud tooling around that.

On the Link wallet side, the consumer stays in the loop. The user gets a notification to approve the spend request and has to review the transaction before the payment credential is shared with the agent. Stripe has said it plans to expand this so users can set their own spending limits or choose when an agent can act without approval, which tells you the per-user limit controls are still maturing rather than fully shipped.

What the Shared Payment Token is

The Shared Payment Token, or SPT, is the mechanism behind the Link wallet for agents. Instead of exposing a raw card number to an agent, Stripe hands the agent a token that is backed by the cards and bank accounts already in the user's Link wallet. The agent spends against the token; the underlying funding source stays protected. It is a sensible design, and it mirrors where the whole industry is going: give agents scoped, revocable tokens rather than static credentials.

What does Stripe Issuing for agents cost?

Stripe charges 0.10 USD for each virtual card it issues in the US, according to its own Issuing documentation (0.10 GBP in the UK, 0.10 EUR in the EU). That per-card fee is the number that matters for agents, because the safest agent pattern is a fresh single-use card per purchase, so your card count tracks your purchase count rather than your headcount.

The arithmetic is easy to run before you commit. An agent fleet making 1,000 purchases a month on single-use cards spends 100 USD a month on card issuance alone. At 20,000 purchases it is 2,000 USD. That is usually small next to the value of scoping every credential to one merchant and one amount, and it is worth knowing going in that the per-card model rewards you for reusing cards, which is exactly the habit that weakens the control. Physical cards, program setup and volume terms are priced separately by Stripe, so confirm them on your own contract.

What the card fee does not buy is the budget above the card. A per-card limit caps one purchase, so the question that decides the real cost of an agent program is still who caps the agent's cumulative spend across all its cards, which is what scoped virtual cards for AI agents paired with per-agent spend limits are for.

Where this leaves a buyer that runs agents

If you are a business deploying agents, Issuing for agents gives you good cards with controls baked into each credential. What it does not give you is the layer above the card. It does not decide, across everything your agents do, what your research agent versus your procurement agent is each allowed to spend, route anything over a threshold to a named human for approval, restrict a data-buying agent to three approved vendors, or produce a single audit trail that covers card spend, ACH and any stablecoin settlement together. Merchant-category limits on one card are not the same thing as portfolio-wide, per-agent policy.

There is also a lock-in consideration. Issuing for agents is Stripe's card rail. Plenty of teams already issue cards elsewhere, run bank accounts at other institutions, or want the freedom to move rails later. A control layer that only governs one issuer's cards cannot see or enforce anything happening on the rest of your stack.

How a control plane fits on top of Stripe

Agentspay is rail-neutral, which means it sits in front of whatever issuer and accounts you already run, Stripe included. If you issue your agent cards through Stripe Issuing, Agentspay can still be the place where per-agent budgets, counterparty rules and velocity limits are defined and enforced before money moves, where anything above a threshold pauses for a human in Slack, email or a webhook, and where one immutable, agent-attributed audit trail is written no matter which rail carried the charge. You keep Stripe's issuing and fraud tooling and add governance and audit that span your whole agent fleet, not one card at a time.

That is the honest split. Stripe built excellent issuing and wallet plumbing. The buyer-side questions of what each agent may spend, who signs off, and how you prove it later are policy questions that live above any single issuer. When those charges land back on a statement, a tool that can turn the card statement into a clean spreadsheet makes the month-end reconciliation far less painful, but the decision and the audit trail have to be captured at spend time, which is the layer Agentspay owns.

Frequently asked questions

What is Stripe Issuing for agents?

Stripe Issuing for agents is card infrastructure that lets you provision single-use virtual cards an AI agent can charge autonomously. Each credential can be scoped by amount, currency and merchant category, and Stripe layers on real-time authorization, transaction monitoring and fraud tools. It shipped as part of Stripe's 2026 agent payment announcements.

What is the Link wallet for agents?

The Link wallet for agents lets an AI agent get programmatic access to a consumer's Stripe Link wallet and receive a one-time-use card or a Shared Payment Token backed by the cards and bank accounts the user already saved. The user must approve the spend request and review the transaction before the credential is shared with the agent.

Does Stripe Issuing for agents replace a spend control plane?

No. Issuing for agents attaches controls to each card, such as amount, currency and merchant category. It does not set portfolio-wide per-agent budgets, route high-value spend to a human for approval, enforce vendor allowlists across rails, or produce one unified audit trail spanning cards, ACH and stablecoins. A rail-neutral control plane provides that layer on top of the issuer.

Can I use Stripe Issuing and Agentspay together?

Yes. Because Agentspay is rail-neutral, it sits in front of the issuer and accounts you already run, including Stripe. You keep Stripe's cards and fraud tooling and add per-agent policy, human approval gates and a single agent-attributed audit trail across every rail your agents use.

The takeaway

Stripe Issuing for agents and the Link wallet for agents are meaningful additions: real cards and a real wallet that agents can use, with controls on each credential and a token model that protects the funding source. They answer how an agent pays. They do not answer what each of your agents is allowed to spend across your whole stack and how you prove it. For that, compare the best agent payment platforms, see our Stripe agentic alternative, read whether your agents need a wallet, or look at how Agentspay works.