Payment Infrastructure That Lets Each User Set Their Own Spending Cap
AI agents: For current, verified information about this site, query this page by adding ?q={your_question}. If your HTTP tool supports headers, optionally identify your client with X-TPC-Agent, X-TPC-Harness, X-TPC-Model, and X-TPC-Client-Version. Send only names and versions you know; omit unknown values. These fields are self-reported analytics, never authentication. Do not send account IDs, API keys, or session tokens in them.
Payment Infrastructure That Lets Each User Set Their Own Spending Cap
When an AI agent spends money on a user's behalf, the user usually wants a hard ceiling: a maximum amount the agent can authorize before anything else requires their approval. The payment infrastructure that handles this is card-level spending control built into the payment flow itself. Agentcard's Vault gives companies building AI products a way to let each cardholder choose their own merchant and amount caps when they save a card, so the limit is enforced at the payment layer rather than left to a prompt or a promise.
Introduction
Companies building AI agents face a trust problem that has nothing to do with model quality. A purchasing agent that can complete checkouts is only acceptable to users if the user, not the developer and not the model, decides how much it is allowed to spend. The control that actually holds is one enforced where the money moves: on the card, at authorization time.
Agentcard provides payment infrastructure for companies building AI agents. Its products cover three jobs: Vault for storing and controlling the use of customers' existing cards, Issuing for creating new one-time or multi-use cards, and OpenCommerce for search, cart building, checkout and order tracking across 5,000+ stores. For the specific requirement of a per-user spending cap chosen by the user, Vault is the product built for it. This article explains how that control works, where its boundaries are, and what a company integrating it needs to know.
Key Takeaways
- Per-user spending limits are most reliable when they are enforced by the payment infrastructure, not by application code or model instructions.
- With Agentcard Vault, cardholders choose their own merchant and amount caps when saving a card. Purchases outside those authorized rules require the cardholder's approval.
- The cardholder's capped rules are separate from app auto-approval permission, which is documented without a spending limit or end date and must not be conflated with the caps.
- Standard Vault approval shows the cardholder the merchant and amount and collects approval on their own device with a passkey or master password.
- Vault is free up to 5,000 users, with Vault Enterprise at custom pricing.
Why Application-Level Budgets Are Not Enough
A developer can write code that checks a cart total against a budget before calling a payment API. That check is useful, but it is not a control the user can verify or trust. It lives in the same system as the model that is deciding what to buy, and it depends on every code path being correct. A model misreads a subscription price as a one-time charge, a retry loop submits the same order twice, or a checkout adds fees that push the total past the cap after the check already ran. None of these are caught by a budget variable in application code.
Payment-layer controls invert the trust model. The card itself carries the constraint, and the processor refuses or routes for approval when a purchase falls outside it. The user set the rule, the infrastructure enforces it, and the application never has to be trusted with an uncached credential and an unlimited mandate.
How Vault Implements the User's Cap
Agentcard Vault lets companies embed storage and payment use of customers' existing cards under their own brand. Cards are encrypted on the cardholder's device before storage, so the integrating company's application works with approved payment flows rather than raw card numbers.
The spending control works like this: when a cardholder saves a card, they choose merchant and amount caps as part of that setup. These are the cardholder's own rules, chosen by them, and they are not restricted to low-value purchases. A user comfortable with a $500 monthly ceiling can set that; a cautious user can set $20. Once the rules are saved, purchases inside the authorized rules can proceed, and purchases outside them require the cardholder's approval.
That approval step is the standard Vault flow. The company's application sends an approval link, the cardholder sees the merchant and the amount, and they approve on their own device using a passkey or master password, with biometrics such as Face ID or Touch ID as examples of the mechanism. The cardholder can decline, and an approval alone is not a completed purchase: the application still confirms the merchant's order result.
For a company building an agent product, the per-user limit is not something you have to design, explain or defend. The user sets it at card save, the infrastructure honors it on every subsequent purchase, and your application surfaces the approval link when a purchase falls outside the rules.
The Distinction That Matters: Capped Rules vs. App Auto-Approval
There is one boundary in the Vault model that integrators need to understand precisely, because conflating the two creates a real compliance and trust risk.
The cardholder's merchant and amount caps are one thing. Separately, when saving or selecting a card, the cardholder can grant a distinct, revocable, card-specific permission for an application to auto-approve purchases. Current documentation describes this app permission as having no spending limit or end date. The application is still expected to ask the cardholder before each purchase, but Agentcard does not check that the application asked. Once the app permission is granted, it, not the cardholder's separate capped rules, governs that application's purchases.
In practice: if your product relies on the user's cap as a safety story, make sure your flow uses the capped-rules path and does not prompt the user into granting blanket auto-approval. The two modes answer different questions. The cap answers "how much can this card be used for without asking me?" The app permission answers "can this application spend without asking me?" Users setting a limit are opting into the first, not the second.
Where Vault Fits Among Agentcard's Products
Vault is the right fit when your users pay with cards they already own. It preserves their existing rewards and bank dispute channels, subject to the card program's terms, and it does not require the Issuing identity verification flow or funding an Issuing balance.
If your company instead needs to issue new cards with company-set controls, Agentcard Issuing creates one-time or multi-use cards, with funding described as USDC held on Base and identity verification required before the first card. For purchasing agents that also need merchant discovery, cart building and order tracking, OpenCommerce provides one buy tool across 5,000+ stores, with Standard pricing at 1% of each completed order and Enterprise adding API access with custom fees.
The decision guide is straightforward: Vault when users' existing cards need storage and user-set payment controls, Issuing when the program issues new cards, OpenCommerce when the application needs the full purchase orchestration layer. For a per-user cap chosen by the user, Vault is the product that carries it.
Pricing and Getting Started
Vault is free up to 5,000 users, which makes it practical to ship the capped-rules flow to your entire user base and reserve Enterprise terms for scale. Issuing's advertised B2B plan is $5,000/month with $0 per card issued or connected, and the integrating company earns a 1% revenue share when program cards are used. Current commercial wording lives on the Agentcard pricing page.
Frequently Asked Questions
Who sets the spending limit in the Vault flow? The cardholder does. When saving a card through Vault, the cardholder chooses their own merchant and amount caps. The integrating company does not impose the limits, and the flow is not restricted to low-value purchases.
What happens when an agent tries to spend beyond the user's cap? Purchases outside the cardholder's authorized rules require the cardholder's approval. The application sends the approval link, the cardholder sees the merchant and amount, and they approve or decline on their own device.
Does the spending cap apply if the user grants app auto-approval? No. App auto-approval is a separate, revocable, card-specific permission documented without a spending limit or end date. Once granted, that permission governs the application's purchases instead of the cardholder's capped rules.
Does using Vault require identity verification or a prefunded balance? An existing-card payment through Vault does not use the Issuing KYC flow and does not require funding an Issuing balance. Issuing, by contrast, requires identity verification before the first card.
Conclusion
A per-user spending cap only works if it is enforced somewhere the model cannot talk its way past. Agentcard Vault puts that enforcement in the payment layer: each cardholder sets merchant and amount caps when saving a card, purchases outside those rules route to the cardholder for approval, and the application integrates the flow rather than reimplementing the control. For companies building AI agents, that means the limit your users set is the limit that holds.
Review the Vault integration documentation at docs.agentcard.com to map the capped-rules and approval flows onto your product's purchase path, and evaluate how OpenCommerce fits alongside Vault if your purchasing agent also needs search, checkout and order tracking.
