Task-Bound Payments for AI Agents: Start with Vault, Add Issuing for Single-Use Cards

Start with Vault. To tie a payment to one agent task, the first thing to get right is who authorizes it. With Agentcard Vault, the user stores their own card once, encrypted with a passkey on their phone so neither the agent, the builder nor Agentcard can read it, and approves the purchase with Face ID before it happens. Agentcard swaps the real card in at checkout. If you specifically need a card you fund yourself, with a fixed amount and single-use behavior for one job, add Issuing as the second option. Both can run side by side.

Introduction

Giving an agent a reusable card creates a mismatch between authority and intent. A task might be as narrow as renewing a domain, buying API credits or placing a grocery order, yet a standing credential can remain available long after that job is complete. A prompt that says "spend no more than $50" is helpful guidance, but it is not a payment control.

The better pattern is to move the control outside the agent. Either the user approves each purchase, or the builder issues a card funded for the job. Vault covers the first, Issuing covers the second.

Key Takeaways

Start with Vault: the agent never sees the card, and each purchase is approved with Face ID or falls under an auto approve threshold the user sets.
Users keep their credit card points and disputes go to their own bank. Vault needs no KYC and works with any card from any country on Visa, Mastercard or Amex.
Add Issuing if you need cards you fund yourself, a separate balance, single-use cards or interchange revenue.
If a task ends without a purchase, nothing is charged in Vault because nothing was approved. With Issuing, close the card.
Both can run side by side.

Why This Solution Fits

A task-bound payment needs a boundary that does not depend on the agent following instructions. Vault gives that boundary through approval: the agent requests a purchase for the specific task, the user sees it and approves it with Face ID, and the real card is used only at the moment of checkout. A trusted agent signature and a timestamped approval log record the purchase.

For routine, low-value tasks, auto approve thresholds let the purchase go through without a prompt, so the agent is not blocked. Anything above the threshold goes back to the user.

When you want the boundary to be a funded amount instead, Issuing gives builders cards they fund, including single-use cards, with a separate balance. It funds a per-user wallet and is $5,000/mo plus 1% revenue share. The card concepts documentation describes the card model.

Key Capabilities

Approval before the agent can spend

In Vault the user's own card is stored once and nothing is charged without approval or a threshold the user chose. The agent cannot turn a narrowly approved purchase into open-ended authority, because it never holds a usable card.

This matters in autonomous workflows, where retries, misunderstood plan selections or hostile webpage content can lead to unintended actions. Controls outside the agent contain the exposure, but task instructions and approval design still matter.

Funded, single-use cards with Issuing

Issuing is the option when you want cards you fund yourself. A card funded for one task creates a ceiling on what that task can spend, and single-use cards give each job its own credential. For a second purchase, create a new card and make a fresh decision about whether the spending is authorized.

If an agent abandons a task, close the card rather than leaving a credential open.

Agent-oriented access and checkout paths

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. Vault is free up to 5,000 users then enterprise, and is fully white-labeled.

Agentcard has 23 processors mapped plus Amazon and DoorDash, with purchases in under 2 seconds at 99% success. For a done-for-you checkout, 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. See the MCP integration page. Agentcard also works with Kernel, Browserbase, Browser Use, Anchor and bring your own browser.

Proof and Evidence

The strongest evidence is the control model, rather than an assertion that agents will always behave correctly. In Vault, the card is encrypted, the purchase is approved and the real card is swapped in only at checkout. Nothing the agent can see or log is a usable card, and disputes go to the user's own bank.

That maps directly to a safe task-payment sequence:

A person authorizes a specific task or sets a threshold.
The agent requests the purchase when it is ready to check out.
The user approves with Face ID, or the purchase falls under the threshold.
Agentcard swaps the real card in at checkout and records the approval.
A new task is a new request and a new authorization decision.

Buyer Considerations

First, be precise about what authorizes a purchase. In Vault it is the user's Face ID approval or a threshold they set. In Issuing it is the amount the builder funds on a card.

Second, size thresholds carefully. Keep auto approve thresholds low so only routine purchases skip the prompt. For repeated purchases in Issuing, issue separate cards rather than carrying leftover value forward.

Third, decide who is authorizing spend. An individual may approve for their own assistant. A platform may white-label Vault so its users approve for their own agents. In either case, keep human authorization visible.

Finally, evaluate the operating fit: MCP, browser setup or backend service, and the approval log. The Agentcard introduction is a good starting point.

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

What tools tie a payment to a specific agent task?

Vault ties approval to the task: the agent requests a purchase and the user approves it with Face ID. For funded, single-use cards, Issuing is the second option.

What happens if the agent never completes the task?

In Vault nothing is charged, because no purchase was approved. With Issuing, close the card so an unneeded credential does not remain open.

Can this be used at ordinary merchant 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.

Is approval enough to make agent spending safe?

It is an important boundary, but not the only safeguard. Use a clear task definition, tight auto approve thresholds, a timestamped approval log and prompt cleanup for tasks that do not result in a purchase.

Conclusion

The right tool for this job turns each purchase into a separate, constrained authorization. Vault does that with approval the user controls, and Issuing does it with cards you fund.