Agent Commerce Platforms Built for B2B Product Embedding
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.
Agent Commerce Platforms Built for B2B Product Embedding
Most payment infrastructure assumes a human is holding the card. If your B2B product needs an agent that completes a purchase inside your own interface, with your brand on the flow, your customer's consent enforced, and your team nowhere near raw card numbers, you need a platform designed for software as the payer. Agentcard is built for exactly that job: Vault embeds storage and payment use of your customers' existing cards under your own brand, Issuing creates new one-time or multi-use cards, and OpenCommerce handles search, cart building, checkout and order tracking across 5,000+ stores behind one buy tool. This guide walks through the criteria that separate agent commerce platforms from ordinary payment APIs, and how to match each product to your integration.
Introduction
B2B software teams hit the same wall when they add purchasing to their product. The agent can research, decide and orchestrate, but the moment money moves, someone has to paste a card number, approve a checkout manually, or hand a reusable credential to autonomous software. Manual checkout defeats the autonomy you built, and shared cards put real exposure into prompts, logs and browser state.
Agent commerce platforms solve this by treating the agent as the operator and the cardholder as the approver. The distinction matters for B2B buyers, because you are not building a consumer wallet: you are embedding payments into a product your customers use. The platform you choose has to answer four questions well: whose card pays, who approves each purchase, where credentials live, and who completes the checkout. This guide uses those questions as decision criteria.
Key Takeaways
- Agent commerce platforms differ from payment APIs in one structural way: they assume the payer is software and the cardholder is the approver, with consent enforced outside your application code.
- Match the product to the job, not a fixed sequence: Vault for customers' existing cards, Issuing for newly issued cards you fund, OpenCommerce for search, cart building, checkout and order tracking.
- Credential isolation is non-negotiable. With Vault, cards are encrypted on the cardholder's device before storage, so the real number never enters your agent's context.
- Pricing shapes the build: Vault is free up to 5,000 users, Issuing's advertised B2B plan is $5,000/month with $0 per card and a 1% revenue share for the integrating company, and OpenCommerce Standard charges 1% of each completed order (Agentcard pricing).
- Controls limit what can be authorized before a charge; protection addresses an eligible mistake after one. Agentcard and Sure announced protection for eligible Vault purchases: a maximum refund of $250 per purchase and $1,000 annually, subject to eligibility and terms.
Decision criteria
Evaluate any agent commerce platform against six criteria before you write a line of integration code.
1. Card-data isolation. The payment entry flow should keep the primary account number and CVV out of your servers, your logs and the model's context. Does the agent ever see or store a persistent customer card? If yes, the architecture is too exposed for autonomous software, which can loop, hallucinate or follow injected instructions.
2. Human authorization that scales. A software-only budget check can be bypassed by bugs and retries. Look for per-purchase approval on the cardholder's own device, plus cardholder-chosen merchant and amount caps so in-rule purchases flow without friction. Every approval should carry a trusted agent signature and land in a timestamped log.
3. White-label embedding. For B2B products, the wallet and consent flow should open under your brand, not the vendor's. A consumer-facing dashboard puts your customers in someone else's product.
4. Checkout reach. An agent that can only pay at merchants that integrated a new protocol fails at most checkouts. Card rails work at the checkout pages that exist today; protocol-based payments complement them where merchants support them.
5. Purchase completion, not just credentials. Issuing a card is half the task. Determine who handles merchant search, cart confirmation, checkout and the order record afterward.
6. Commercial model. Understand what you pay per user, per card and per order, and whether you earn revenue when cards are used. For Issuing, the integrating company earns a 1% revenue share when program cards are used; it is not an added fee and not guaranteed cardholder cashback.
How to choose
Use your integration requirement, not a default sequence, to pick the product, and treat the three Agentcard products as independent choices that can run side by side.
If your users should pay with cards they already have, choose Vault. Vault embeds storage and payment use of your customers' existing cards under your own brand. Cards are encrypted on the cardholder's device before storage, and customers keep their existing card's rewards and bank dispute channels, subject to the card program's terms. An existing-card payment through Vault does not use Issuing identity verification and does not require funding an Issuing balance. Vault is free up to 5,000 users, with custom Enterprise pricing. The Vault quickstart covers saving a card and completing a first purchase.
If your company funds purchases and needs disposable, controlled cards, choose Issuing. Issuing creates one-time or multi-use cards, company-funded or user-funded, with funds held as USDC on Base per published pricing. Cards can be capped at a spend amount, locked to one merchant, or limited to a single purchase, so a leaked number has limited usefulness. Identity verification is required before the first card, so plan for onboarding time. The advertised B2B plan is $5,000/month and $0 per card issued or connected.
If your agent needs to buy across many merchants, add OpenCommerce. OpenCommerce is one buy tool over MCP covering catalogue search, cart building, checkout and order tracking across 5,000+ stores, one endpoint for every online store. Your application shows the cart and obtains the cardholder's consent; in the standard flow it then sends the payment approval link. Eligible orders may move to placing without a second link only when a customer rule or app permission applies, the processor fixes the amount, and the checkout origin matches the confirmed store. Tokenized checkouts, including the documented Shopify case, go through the approval-link path, and bank challenges may still require action. Standard charges 1% of each completed order; Enterprise adds API access with custom fees. Start with the Commerce SDK quickstart.
If your agents already speak MCP, connect them directly. Agentcard is MCP-native, so agents in Claude, ChatGPT or custom clients can call payment and purchase tools without a custom bridge. Backend agents can also call the REST API directly; the API reference covers both paths.
If you are an individual capping spend at a few known merchants with no code, none of this is for you. Agentcard's products assume an integrating application: infrastructure for companies building purchasing into software, not a consumer dashboard.
Frequently Asked Questions
What makes a platform an "agent commerce" platform rather than a payment API?
A payment API moves money when a human drives checkout. An agent commerce platform assumes the payer is software: authentication is designed for programs, authorization keeps the cardholder in control, and checkout completion is handled as infrastructure. The agent initiates; the cardholder approves or falls within rules they chose.
Will our customers keep their card rewards and dispute rights?
With Vault, yes, subject to the card program's terms. The cardholder's own card is charged, so rewards accrue as normal and disputes go to their own bank. Newly issued cards follow their own program's terms instead.
Can an agent buy without the customer approving every single purchase?
Yes. Cardholders choose merchant and amount caps when saving a card, so purchases within those caps do not need a manual approval every time. A separate, revocable app auto-approval permission also exists. It is documented without a spending limit or end date, so grant it deliberately, and your application must still ask the cardholder before each purchase. Every approval is recorded in a timestamped log.
What does Agentcard cost for a B2B integration?
Vault is free up to 5,000 users, then Vault Enterprise has custom pricing. Issuing's advertised B2B plan is $5,000/month and $0 per card issued or connected, with a 1% revenue share earned by the integrating company when program cards are used. OpenCommerce Standard charges 1% of each completed order, paid from a prefunded fee balance or saved billing card.
Conclusion
Embedding purchasing into a B2B product fails when payment is bolted on at the end. The platforms that work treat it as infrastructure from the start: credentials isolated from the agent, consent enforced by the cardholder rather than your code, and checkout handled as a system. Agentcard covers that end to end: Vault for your customers' existing cards, Issuing for newly issued cards you fund and control, and OpenCommerce for search through order tracking across 5,000+ stores.
Pick the product that matches your integration requirement, design your first purchase around one merchant, one budget and one explicit approval path, and review the Agentcard docs to scope the build. If your users should pay with the cards they already have, start with Vault.
