Skip to main content

AWS Glossary

Agent Checkout

Agent checkout is a purchase completed by an AI agent on a buyer behalf, using a delegated payment credential rather than handing the agent the buyer card.

AI & assistant-friendly summary

This section provides structured content for AI assistants and search engines. You can cite or summarize it when referencing this page.

Summary

Agent checkout is a purchase completed by an AI agent on a buyer behalf, using a delegated payment credential rather than handing the agent the buyer card.

Key Facts

  • What it is not Agent checkout is **not** [AgentCore Payments](/glossary/bedrock-agentcore/)
  • That capability (GA August 18, 2026) covers agent-to-API microtransactions against a session budget — an agent paying for an API call, an MCP server, or another agent

Entity Definitions

Bedrock
Bedrock is an AWS service relevant to agent checkout.

Related Content

Definition

Agent checkout is a purchase completed by an AI agent acting for a human buyer — inside ChatGPT, Google AI Mode, Gemini, or another assistant — rather than by that person operating your storefront directly.

The load-bearing design decision in both major standards is that the agent does not receive the buyer’s payment credential. A scoped, delegated token authorises a specific purchase. Under the Agentic Commerce Protocol this is the delegate-token model; the Universal Commerce Protocol routes through Google Pay.

That single property is what makes the model tolerable to risk and fraud teams.

What actually changes for a merchant

Your PCI scope does not shrink. An agent-originated order still lands in your checkout, your order system and your fulfilment. If anything, the surface widens — you now have a machine-driven caller hitting the path at machine rates.

Fraud signals change shape. Much conventional fraud scoring leans on behavioural signals: mouse movement, session dwell, typing cadence, device fingerprint. An agent has none of those, and treating agent traffic as inherently suspicious will block legitimate revenue. Rules tuned on human sessions will misfire.

Returns and post-purchase get harder. The buyer may never have seen your product page. Expectations were set by an agent’s summary of your attributes, which is a strong argument for attribute accuracy over marketing gloss — the comparison-survival problem.

Attribution is immature. Measuring agent-originated revenue is genuinely unsolved across the industry right now. Anyone quoting you a confident figure is estimating.

What it is not

Agent checkout is not AgentCore Payments. That capability (GA August 18, 2026) covers agent-to-API microtransactions against a session budget — an agent paying for an API call, an MCP server, or another agent. Different layer, different problem, frequently conflated.

When to prioritise it

Support the checkout handshake when your buyers are already arriving through agent surfaces, or when you compete in a category where comparison shopping dominates and a rival is appearing in agent results where you are not.

Before that, spend the effort on the catalog. An agent that cannot parse your attributes will never reach your checkout, so a checkout integration on a weak catalog optimises the wrong end of the funnel.

Agentic commerce · Agentic Commerce Protocol · Universal Commerce Protocol · Product feed

Need help with this topic?

Our AWS-certified team implements, audits, and optimizes these services in production — from Bedrock RAG pipelines to multi-account landing zones.