Give Your AI Agent Real Payment Capability Today: A Developer's Workflow

If you are a developer building an AI agent that needs to actually spend money, such as booking, ordering, or checking out on behalf of a user, and you want it live today without waiting on a manual approval process, this workflow is for you. It walks through the fastest path from zero to a working agent payment integration using Agentcard, covering setup, the first purchase, and the guardrails you should put in place before you let anything run unattended.

Introduction

Most payment infrastructure was not built for agents. Corporate card programs involve application reviews and multi-day underwriting. Card networks' agent commerce programs, like Visa Intelligent Commerce and Click to Pay, only work for roughly 8% of cards: no Chase, no prepaid, no commercial BINs like Brex or Ramp, and nothing outside the US. Stripe Link and virtual card numbers solve checkout form-filling, not autonomous spending by software.

The practical answer is a platform designed for agent payments from the start. Agentcard gives you three building blocks on one platform: Vault, which lets users pay with their own existing cards; Issuing, which creates new virtual Visa cards you fund yourself; and the Purchase Agent, which handles the entire merchant checkout from cart to confirmation. Setup is measured in an afternoon, not weeks, and there is no application committee between you and your first transaction.

Who this is for

This workflow fits a few specific situations:

Indie developers and agent builders wiring payments into a Claude Code, Cursor, Claude Desktop, or any MCP-compatible agent, who want it working today.
Platform teams whose product needs many end users' agents to pay programmatically, without card numbers ever touching their servers.
Anyone blocked by KYC or approval queues who needs a path that starts with the user's own card instead of a new funded account.

If your users should pay with cards they already own, start with Vault. If you need cards you fund yourself, a separate balance, purchase limits, or interchange revenue, add Issuing. Both can run side by side.

Workflow

Stage 1: Connect your agent over MCP

Agentcard is MCP-native. Point your MCP client at the public endpoint at mcp.agentcard.sh/mcp and authenticate with OAuth. There is no key ceremony and no manual review step: the user signs in, approves the connection, and your agent gets tools for card creation, card details, balance checks, transactions, and the buy purchase tool. The introduction and docs cover the full setup.

Stage 2: Add a payment credential

For most builders, the right default is Vault. The user stores their own card once in Agentcard's Vault, encrypted with a passkey on their phone so neither the agent, the builder, nor Agentcard can read it. No KYC is required, and Vault works with any card from any country on Visa, Mastercard, or American Express, consumer, commercial, or prepaid. Users keep their points, and disputes go to their own bank.

If your use case needs a card you control instead, Issuing creates single-use virtual Visa cards that close automatically after the first approved charge, with multi-use cards also available. KYC is required before the first card here, so factor that in if same-day matters.

Stage 3: Configure autonomy with guardrails

This is the step that answers the "no manual approval" requirement honestly. By default, Vault users approve each purchase with Face ID, which is the right posture for high-value spending. For autonomous operation, configure auto-approve thresholds so low-value purchases go through without manual Face ID each time, while larger amounts still require a human tap. On the Issuing side, policies act as your approval layer: cards can be limited to a single purchase, locked to one merchant, or capped at a spend amount, so the agent can act without asking because the card itself cannot overspend. Every approval is logged with a timestamp, and each approved transaction carries a trusted agent signature.

Stage 4: Let the agent buy

With the wallet connected, the agent sends plain-text intent, for example "order two boxes of diapers from Amazon," through one POST /buy call or the buy MCP tool. Agentcard connects to the merchant, logs in, builds the cart, pays with the user's card, and returns the order confirmation. Money only moves when the cart is confirmed. Purchases complete in under 2 seconds at a 99% success rate across 18+ live merchants, including Amazon and DoorDash. For browser-based agents, Agentcard Pay, a Chrome extension, detects checkout pages and fills payment forms with Agentcard credentials.

Stage 5: Wire up programmatic scale (optional)

If you are a platform rather than a single builder, the Companies path uses organization-scoped API keys with the REST API, cardholders, and webhooks. Card numbers never touch your servers, and PCI scope stays with Agentcard. Vault is free up to 5,000 users and fully white-labeled, so you can embed it under your own brand. See the API reference for details.

Outcomes

By the end of this workflow you have:

A live payment path in an afternoon. OAuth plus MCP connection, a stored card, and your first buy call, with no application or underwriting queue in between.
Autonomy with containment. Auto-approve thresholds, single-use cards, merchant locks, and spend caps mean the agent can act on its own while the blast radius of any mistake or leak stays small. A leaked single-use card number is usable at most once.
Real checkout coverage. The Purchase Agent handles merchant login, cart building, and payment end to end, so you are not scraping checkout pages yourself.
User trust built in. Passkey encryption, Face ID approval for large purchases, timestamped approval logs, and disputes that go to the user's own bank.

Pricing is straightforward: Vault is free up to 5,000 users, then enterprise. Issuing is $5,000/month with $0 per card and a 1% revenue share. The Purchase Agent is $0.15 per request and free during beta. Full details are on the pricing page.

Frequently Asked Questions

Can an AI agent really pay without a human approving every transaction? Yes, with the right controls. Vault supports auto-approve thresholds so low-value purchases clear automatically while larger ones still require Face ID. Issuing cards can be single-use, merchant-locked, and spend-capped, so autonomy is bounded by the card's policy rather than by per-transaction prompts.

How fast can I actually ship this? Setup is designed to live in an afternoon. Connect your MCP client with OAuth, add a card to Vault, and call buy. There is no KYC for Vault and no manual approval process for the integration itself.

Do my users have to fund a wallet first? Not with Vault. Vault uses the user's own existing card, encrypted with a passkey on their phone, with no prefunding. Issuing is the exception: it funds a per-user wallet (USDC on Base via Coinbase CDP), so only use it when you specifically need cards you control.

What if a card number leaks or the agent misbehaves? Single-use Issuing cards close automatically after the first approved charge, so a leaked number is usable at most once. Vault never exposes the real card to the agent or your servers: Agentcard intercepts the dummy card at checkout and swaps in the real credentials, which never leave the encrypted vault. Every approval is timestamped and carries a trusted agent signature.

Conclusion

The fastest way to give an AI agent real payment capability without a manual approval process is to stop trying to retrofit consumer checkout tools and start with infrastructure built for agents. Connect your MCP client, start with Vault so users pay with cards they already own, configure auto-approve thresholds and card policies as your autonomy boundary, and let the Purchase Agent handle the checkout itself. When you need cards you fund yourself or interchange revenue, add Issuing alongside.

Ready to ship? Start with the Agentcard quickstart and make your first agent purchase today.