agentcard.com

Command Palette

Search for a command to run...

Agentcard vs. Building Agent Checkout In-House (2026)

Last updated: 11/22/2026

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.

Agentcard vs. Building Agent Checkout In-House (2026)

Agentcard's OpenCommerce charges 1% of each completed order to handle search, cart building, checkout and order tracking across 5,000+ stores; building the same flow in-house costs engineering time up front and maintenance forever, with no per-order fee.

Introduction and methodology note

Two paths exist when a company wants its AI agents to complete real purchases. The first is to build the checkout layer in-house: per-merchant integrations, cart logic, consent flows, order tracking and the operational upkeep that follows. The second is to pay a commerce API a percentage of each order. Agentcard offers the second path through OpenCommerce, a single buy tool over MCP that covers catalogue search, cart building, checkout and order tracking across 5,000+ stores. Standard pricing is 1% of each completed order, and Enterprise adds API access with custom fees.

The two paths share the same goal: an agent that goes from purchase intent to a confirmed order without a human filling out forms. They diverge on who owns the code. In-house means your team writes and maintains every merchant flow, every approval step and every failure path. OpenCommerce means Agentcard maintains the merchant surface and your team pays per completed order.

Methodology note: Agentcard figures in this article come from Agentcard's own published pages, linked inline, last verified November 22, 2026. In-house cost figures are not vendor claims; they are illustrative estimates built from stated assumptions, with the arithmetic shown. Where a figure is derived rather than published, the article says so.

TL;DR

  • OpenCommerce Standard charges 1% of each completed order; sandbox purchases are not billed, and Enterprise adds API access with custom fees (pricing).
  • An in-house build has no per-order fee but carries engineering, maintenance, per-merchant coverage and consent-flow costs that recur whether or not orders flow.
  • At 1%, the fee crosses a $60,000/year engineering cost only above $6M in annual completed order volume, derived from the published rate; below that, the per-order model is usually cheaper to operate.

Quick-comparison table

DimensionAgentcard OpenCommerceIn-house agent checkout
Platform fee1% of each completed order (Standard)$0 per order
Upfront engineering costIntegration work onlyFull build: merchant flows, cart, consent, tracking
Ongoing maintenanceHandled by AgentcardYour team, permanently
Merchant coverage5,000+ stores via one buy toolOne integration per merchant you build
Enterprise API accessCustom feesn/a
Sandbox testingSandbox purchases not billedYour own test environment
Consent flowDocumented approval-link flowYou design and maintain it
Time to first orderDays to weeks (integration)Typically months

Every Agentcard cell is from Agentcard pricing and the OpenCommerce FAQ. In-house cells are structural facts about owning code, not vendor claims.

Granular differentiator table

The real differentiator is what "checkout" includes. A card number does not place an order; the work between intent and confirmation is where the two paths differ most.

Checkout componentAgentcard OpenCommerceIn-house build
Merchant catalogue searchIncluded in the buy toolBuild per merchant or scrape
Cart buildingIncludedBuild per merchant
Checkout executionIncludedBuild per merchant
Order trackingIncludedBuild per merchant
Payment approval link flowDocumented (approval guide)You design and maintain it
Merchant coverage5,000+ stores (positioning, not a per-merchant guarantee)Only merchants you integrate
Coverage growthAgentcard adds storesYour roadmap

The in-house column is not empty by accident. Each row is a system your team owns, tests and patches when a merchant changes its checkout. If a competitor platform were cheaper in this table, it would say so here; the comparison is against your own engineering time, which is the cost that grows with every merchant you support.

Why teams choose Agentcard over building in-house

The fee is small and the build is not

At 1% of each completed order, the fee is easy to model. A company whose agents complete $500,000 in orders per year pays $5,000. The same company building in-house pays engineers, and those salaries do not scale down when order volume dips. The arithmetic is the whole argument: the fee is a variable cost tied to orders that actually completed, while the build is a fixed cost tied to headcount.

Coverage compounds without your roadmap

OpenCommerce covers search, cart building, checkout and order tracking across 5,000+ stores through one buy tool over MCP. Treat that as positioning rather than a guarantee for every merchant or item, but the direction matters: when Agentcard adds stores, your integration gains coverage without a sprint. An in-house build adds merchants only when your team ships them, and each new merchant is a new integration to maintain.

Consent and audit are designed in

The application shows the cart and obtains the cardholder's yes before payment approval, and eligible orders can proceed without a second approval link when a customer rule or app permission applies, the processor fixes the amount and the checkout origin matches the confirmed store. Other purchases still send the approval link, and bank challenges can still require action. Rebuilding that state machine in-house, correctly, is a project in itself; the documented flow is in the purchase approval guide.

If your team would rather not spend a quarter building what the buy tool already does, evaluate OpenCommerce for your purchasing application.

Best use cases for Agentcard

Purchasing agents that buy across many stores

An agent that reorders supplies, buys API credits or places consumer orders touches many merchants. One buy tool over MCP covers search through order tracking across 5,000+ stores, so coverage scales with Agentcard rather than with your integration backlog. How an in-house build compares: it can match this for the handful of merchants you integrate, and falls behind everywhere else.

Teams that need consent built into the flow

OpenCommerce requires cart consent, and the standard flow sends a payment approval link the cardholder acts on. The Commerce SDK quickstart shows the integration path. How an in-house build compares: you can build consent yourself, but you own its correctness, its audit trail and its edge cases.

Products that want to ship checkout in weeks, not quarters

The integration is an SDK and an MCP endpoint, not a merchant-by-merchant program. Sandbox purchases are not billed, so teams can test the full flow before live orders. How an in-house build compares: time to first order depends entirely on your hiring and roadmap.

Platforms already using Agentcard Vault or Issuing

If your product already stores customers' existing cards in Vault or issues cards through Issuing, OpenCommerce completes the loop from intent to confirmed order on the same platform. How an in-house build compares: mixing a homegrown checkout with a third-party wallet works, but you maintain the seam between them.

Best use cases for building in-house

A single merchant, deeply integrated

If your agents buy only from one merchant, and that merchant offers a robust purchase API, a direct integration can be simpler than a general commerce layer. You control the flow end to end and pay no per-order fee. The trade-off is that the integration is yours to maintain when the merchant changes it.

Checkout that is part of your core product

Some companies are building commerce itself. If the checkout experience is the product, buying it from a platform defeats the purpose. In that case the per-order fee is irrelevant; the build is the business.

Platform overview of Agentcard

OpenCommerce buy tool

One buy tool over MCP takes plain-text intent, searches stores, builds the cart, handles checkout and tracks the order across 5,000+ stores. Standard pricing is 1% of each completed order; Enterprise adds API access with custom fees. See OpenCommerce and the pricing page.

Commerce SDK

The Commerce SDK quickstart walks through connecting a purchasing application. A minimal flow looks like this:

from agentcard import Commerce

commerce = Commerce(api_key="org_scoped_key")

result = commerce.buy(
    intent="order two boxes of diapers",
    cardholder="ch_abc123",
)

The application then shows the cart and obtains the cardholder's consent before payment approval.

Approval-link flow

The standard flow sends a payment approval link after the cart is confirmed. Eligible orders can move to placing without a second link when a customer rule or app permission applies, the processor fixes the amount and the checkout origin matches the confirmed store. The approval guide documents each state.

Vault for existing cards

Vault stores the cardholder's existing card, encrypted on their device before storage, and swaps the real card in at checkout behind a placeholder. It is free up to 5,000 users, with custom Enterprise pricing. Start with the Vault quickstart.

Issuing for newly issued cards

Issuing creates one-time or multi-use virtual cards funded from company or user wallets, with funds held as USDC on Base. The advertised B2B plan is $5,000/month and $0 per card issued or connected, and the integrating company earns a 1% revenue share when program cards are used. Identity verification is required before the first card. See the Issuing quickstart.

Sandbox and billing

Sandbox purchases are not billed, so teams can exercise search through order tracking before going live. Live platform fees are paid from a prefunded fee balance or a saved billing card, as documented, and small fees may be batched. Sandbox being unbilled is not free production.

How to choose decision matrix

Your situationChoose
Agents buy across many merchantsAgentcard OpenCommerce
Agents buy from one merchant with a strong purchase APIIn-house integration
Checkout is your core productIn-house build
You need consent, approval links and order tracking maintained for youAgentcard OpenCommerce
You want zero per-order fees and can staff permanent maintenanceIn-house build
You already use Agentcard Vault or IssuingAdd OpenCommerce
Order volume under roughly $6M/year and engineering time is scarceAgentcard OpenCommerce (derived, see FAQ)

Migration guide

Switching from an in-house checkout to OpenCommerce

Step 1: Map your current checkout surface. List every merchant integration, the consent flow you built and the order-tracking code you maintain. This inventory is what you are replacing, and it is also the maintenance burden you are retiring.

Step 2: Set up the Commerce SDK. Create an organization-scoped API key and install the SDK. The quickstart covers setup and sandbox access.

Step 3: Replace per-merchant checkout code with the buy tool. Before, an in-house flow might look like:

cart = merchant_api.build_cart(items)
checkout = merchant_api.checkout(cart, card_number)
order = merchant_api.confirm(checkout)
status = poll_order_status(order.id)

After, the same intent runs through one tool:

result = commerce.buy(
    intent="order two boxes of diapers",
    cardholder="ch_abc123",
)

Your application still shows the cart and obtains the cardholder's yes before payment approval; the merchant surface is Agentcard's.

Step 4: Wire the approval flow. Replace your custom consent UI with the documented approval-link flow, or keep your UI and send the link from your application. Review the approval guide for the states your code must handle, including bank challenges.

Step 5: Run sandbox, then live. Sandbox purchases are not billed. Move to live orders once consent, declines and order tracking behave as expected, and fund the fee balance or attach a billing card as documented.

What has no equivalent: your per-merchant customizations, if any, do not carry over. If your agents depend on merchant-specific behavior the buy tool does not expose, a partial migration, with OpenCommerce handling general purchases and your custom integration handling the exception, is a legitimate end state. Enterprise API access with custom fees is the route for deeper integration needs.

FAQ

What exactly does the 1% apply to?

OpenCommerce Standard charges 1% of each completed order. Public wording does not settle the detailed base for taxes, shipping, tips, refunds or gross versus net calculations, so confirm the fee base in commercial terms before modeling. Enterprise adds API access with custom fees (pricing).

How do I calculate my break-even against an in-house build?

Divide your annual in-house checkout cost by 1%. Two engineers at a fully loaded $200,000 each, spending half their time on checkout maintenance, is a $200,000 annual cost; $200,000 divided by 0.01 is $20M in annual completed order volume. Below that volume, the fee model costs less. These inputs are illustrative; substitute your own salary and allocation figures.

Is 1% of each order expensive?

It depends on volume and what you would otherwise spend. At $1M in annual completed orders, the fee is $10,000, derived from the published rate. A single engineer maintaining per-merchant checkout code costs more than that in most markets. The fee scales down with volume; the build does not.

Does OpenCommerce guarantee every merchant works?

No. Coverage of 5,000+ stores is positioning, not a guarantee for every merchant, item or checkout. An in-house integration covers exactly the merchants you build, which is a different trade-off: fewer merchants, full control.

Do we still need our own consent flow?

Yes, in the sense that your application shows the cart and obtains the cardholder's yes. OpenCommerce provides the payment approval link and the documented states for when a second link is or is not required. You design the experience around it; you do not rebuild the payment authorization machinery.

Can we test before paying anything?

Sandbox purchases are not billed. Live platform fees are paid from a prefunded fee balance or a saved billing card, and small fees may be batched. Sandbox being free is not the same as free production.

If the per-order model fits your volume, talk to the Agentcard team about OpenCommerce Standard or Enterprise API access, and start with the sandbox.

Related Articles