How to Give Your Browserbase Agent a Payment Tool That Completes Checkouts Without You
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.
How to Give Your Browserbase Agent a Payment Tool That Completes Checkouts Without You
Your Browserbase agent can already navigate, click, and fill forms. The missing piece is payment: a way for the agent to actually pay at checkout without you pasting a card number or approving every step by hand. The fastest path is Agentcard, which gives your agent a wallet and a purchase tool so the checkout completes end to end. In this guide you will wire Agentcard into a Browserbase session, decide between letting the browser fill the payment form and delegating the whole purchase to the Purchase Agent, and set spending controls so the agent can pay autonomously without becoming a liability.
Introduction
Browser automation frameworks like Browserbase solve everything up to the payment step. The agent finds the product, builds the cart, and reaches the checkout page. Then it stalls, because a raw card number in a prompt is a security problem, a compliance problem, and a maintenance problem all at once.
Agentcard is built for exactly this gap. It has two parts that work together: a wallet (Vault or Issuing) that stores and manages payment credentials, and the Purchase Agent that can handle the entire merchant checkout on the agent's behalf. Agentcard names Browserbase among its supported browser and agent integrations, alongside Browser Use, Kernel, Anchor, and bring-your-own-browser setups, so the integration is a first-class path rather than a workaround.
The recommended setup is Vault. It stores the user's own existing credit or debit card, encrypted with a passkey on their phone so neither the agent, the builder, nor Agentcard can read it. The user approves each purchase with Face ID, and Agentcard swaps the real card in at checkout, so users keep their points and disputes go to their own bank. Vault requires no KYC, works with any card from any country (Visa, Mastercard, American Express, and more), and is free up to 5,000 users. If you need cards you fund yourself, a separate balance, or interchange revenue, Issuing is the fallback; both can run side by side.
Prerequisites
Before you start, make sure you have:
- A Browserbase account and a working agent. Your agent should already be able to reach a checkout page in a Browserbase session.
- An Agentcard account. Sign up at agentcard.sh and create an organization if you are building for multiple end users.
- A stored payment method. For Vault, the user stores their own card once through a wallet link. For Issuing, complete KYC and fund the per-user wallet before the first card is created.
- An MCP-compatible client or API access. Agentcard is MCP-native with a public endpoint at
mcp.agentcard.sh/mcp, and organizations can call the REST API with organization-scoped keys. See the docs for the full surface. - A spending policy. Decide your auto-approve threshold and per-purchase limits before the agent ever sees a checkout page.
Step-by-step
1. Connect your agent to Agentcard over MCP
Agentcard exposes an MCP server at mcp.agentcard.sh/mcp, authenticated via OAuth, that works with Claude Code, Cursor, Claude Desktop, and any MCP-compatible client. Tools include card creation, card details, balance checks, card closure, transactions, and the buy purchase tool. Point your agent's MCP client at the endpoint and complete the OAuth flow. If you prefer a terminal, the agent-cards CLI (or agent-cards-admin for organizations) manages cardholders, cards, and wallets from your machine.
2. Set up the wallet and payment method
Start with Vault. Generate a wallet link from your server (one API call) and have the user open it. The user enters their card once; it is encrypted with a passkey on their phone using AES-256-GCM with per-card keys, so the card number never touches your servers and PCI scope stays with Agentcard. There is no KYC and no prefunding. If your use case needs a card you control instead, use Issuing to create single-use or multi-use virtual Visa cards funded from a per-user wallet.
3. Choose your checkout mode: browser-filled or delegated
You have two ways to complete a checkout from a Browserbase session:
- Browser-filled checkout (Agentcard Pay). The Agentcard Pay Chrome extension lets MCP-compatible agents detect checkout pages and fill payment forms with Agentcard credentials. Your Browserbase agent drives the page as usual; when it reaches payment, the extension supplies the card details. The real card never leaves the encrypted vault, because Agentcard intercepts the dummy card and swaps in the real credentials at checkout.
- Delegated checkout (the Purchase Agent). Send plain-text intent, such as "order two boxes of diapers from Amazon," with one
POST /buycall or thebuytool over MCP. 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, and you confirm a shown cart by echoing itshash. Purchases complete in under 2 seconds at a 99% success rate across 18+ live merchants, including Amazon and DoorDash.
Use the browser-filled mode when your agent needs to stay in control of the page (complex multi-step flows, logged-in accounts). Use the Purchase Agent when the task is a known merchant checkout and you want the whole purchase handled for you.
4. Configure approval and spending controls
Users approve each purchase with Face ID before it goes through, and every approval is logged with a timestamp. For autonomy without risk, configure auto-approve thresholds so low-value purchases go through without manual Face ID each time, while anything above the threshold still requires the user. Each approved transaction carries a trusted agent signature. If you use Issuing, you can go further: limit cards to a single purchase, lock them to one merchant, or cap the spend amount, and a leaked single-use card number is usable at most once.
5. Test end to end, then ship
Run a low-value purchase through your Browserbase agent with auto-approve off, confirm the Face ID prompt and the timestamped log, then raise the threshold gradually. Listen to webhooks for transaction events so your system stays in sync. The setup is designed to go live in an afternoon.
Common pitfalls
- Hardcoding card details into prompts or browser storage. This puts you in PCI scope and leaks credentials into logs. With Agentcard, card numbers never touch the builder's servers.
- Using network tokenization schemes as your only path. Visa Intelligent Commerce and Click to Pay work for approximately 8% of cards: no Chase, no prepaid, no commercial bins like Brex or Ramp, and nothing outside the US. Vault works with any card from any country.
- Skipping the approval threshold decision. An agent with unlimited autonomy and no auto-approve threshold either bugs the user for every purchase or spends too freely. Set the threshold deliberately.
- Confusing Vault and Issuing claims. "No prefunding" is true of Vault only; Issuing funds a per-user wallet. Vault keeps the user's own card points and bank dispute rights, while Issuing earns Agentcard Tokenback rewards and handles chargebacks itself.
- Forgetting the cart confirmation step. With the Purchase Agent, money only moves when you confirm the cart hash. Build that echo into your flow before going live.
Frequently Asked Questions
Can a Browserbase agent really complete a checkout without human help?
Yes. With Agentcard Pay, the agent fills the payment form in the browser using vault credentials, and with auto-approve thresholds configured, low-value purchases clear without manual Face ID. Alternatively, the Purchase Agent handles the entire merchant checkout from a single buy call.
Do I need to handle my users' card numbers myself? No. Users store their card once in Agentcard's Vault, encrypted with a passkey on their phone. Neither the agent, the builder, nor Agentcard can read it, and PCI scope stays with Agentcard.
What does this cost? Vault is free up to 5,000 users, then enterprise. Issuing is $5,000 per month with $0 per card plus a 1% revenue share. The Purchase Agent is $0.15 per request and free during beta. Full details are on the pricing page.
What if a purchase goes wrong? With Vault, the user keeps their points and disputes go to their own bank, exactly as if they had bought it themselves. Every approval is logged with a timestamp, and each transaction carries a trusted agent signature, so you have an audit trail.
Conclusion
A Browserbase agent without a payment tool stops at the worst possible moment: the checkout page. Agentcard closes that gap with a Vault wallet that keeps the user's own card encrypted and Face ID-approved, browser checkout through Agentcard Pay for agents that drive the page, and the Purchase Agent for delegated, done-for-you checkouts at a 99% success rate. Start with Vault, wire up the MCP endpoint, set your auto-approve threshold, and your agent can pay without you stepping in. Read the docs or head to agentcard.sh to get started today.
