# AWS lets AI agents bill Marketplace purchases to cloud accounts

Published 2026-11-20 · AI.info Team · AI.info

> AI agents can pay per request for eligible AWS Marketplace products through Amazon Bedrock AgentCore payments. Charges settle in US dollars on the developer's AWS bill, subject to session budgets and other controls.

AI agents can now buy eligible AWS Marketplace products per request and charge the purchases in US dollars to their developers' AWS bills, without stablecoin wallets or API keys. The November 19 launch adds AWS Billing as a payment option in [Amazon](https://aws.amazon.com/blogs/machine-learning/enable-ai-agents-to-pay-for-aws-marketplace-x402-products-with-amazon-bedrock-agentcore-payments/) Bedrock AgentCore payments, AWS said in a technical post.

The option covers x402-enabled software, APIs, data and content. Developers can also use an optional AWS Marketplace connector in AgentCore Gateway to let agents discover products and their endpoints.

## From quote to monthly bill

AgentCore payments became generally available in August 2026 for agents paying for APIs, MCP servers and web content through protocols including x402 and MPP. The Marketplace integration extends that payment capability to products listed by participating sellers.

At launch, AWS Billing supports the x402 exact scheme, under which a seller quotes a fixed price for each request. An agent starts by requesting a seller's endpoint without payment; the seller responds with HTTP 402 and a signed quote in the payment-required header.

AWS Marketplace checks the quote's signature against the seller's registered key. The agent passes the quote to AgentCore payments, which checks the session budget and, if funds remain, issues a payment token. The agent then retries the request with the token to receive the content.

Payment tokens are valid for about five minutes. If the budget check fails, the payment call fails and the developer is not charged. Settlement runs asynchronously, and charges appear on the account's monthly AWS bill.

## Controls and account setup

To use the billing option, a developer creates a Payment Manager with an AWS Billing connector and sets up a service-linked role for Marketplace access. For each task, the developer can create a payment session with a spending limit and give its session ID to the agent.

A first purchase from a seller creates an agreement between that seller and the AWS account; later purchases are tracked against it. Each product appears as a Marketplace line item. Organizations using Private Marketplace restrict agents to the products their configuration allows.

Developers can set spending limits and approved sellers through AgentCore policy, while AgentCore payments enforces the session cap. Agent purchases also appear in AWS CloudTrail and AWS Cost Explorer.

## Research-agent example

AWS provides an open-source sample in which a research agent searches Marketplace for a tool, obtains an endpoint, pays for a request and returns an answer with sources. The walkthrough uses [Tavily](https://ai.info/companies/tavily), a live search provider and Marketplace seller.

In the sample, separate roles divide the ability to set budgets from the ability to pay: the session role cannot pay, and the agent role cannot create sessions. Developers can also skip the discovery connector and specify a seller endpoint for the agent to use.

The sample's default session budget is $0.05, with a $0.02 per-request cap. In a test run described by AWS, the agent compared two listings before making one $0.01 purchase from a $0.01 seller.

## Sources

1. Original source · Blog post [Amazon Web Services](https://aws.amazon.com/blogs/machine-learning/enable-ai-agents-to-pay-for-aws-marketplace-x402-products-with-amazon-bedrock-agentcore-payments/) aws.amazon.com

## Companies in this story

- [Amazon](https://ai.info/companies/amazon)
- [Amazon Web Services](https://ai.info/companies/amazon-web-services)

## Related stories

- [AWS details open-source stack for Bedrock WhatsApp agents](https://ai.info/news/aws-open-source-bedrock-whatsapp-agent-stack)
- [AWS adds stalled-job recovery and checks to SageMaker HyperPod](https://ai.info/news/aws-sagemaker-hyperpod-workload-recovery-checks)
- [AWS adds model leaderboard to SageMaker JumpStart](https://ai.info/news/aws-sagemaker-jumpstart-model-leaderboard)
- [AWS adds AI data preparation tools to SageMaker AI](https://ai.info/news/aws-sagemaker-ai-data-preparation-tools)
- [Exa says AWS chip optimizations lifted embedding throughput about 50%](https://ai.info/news/exa-embedding-throughput-aws-ai-chips)

## Explore

[More articles](https://ai.info/news)

---

Canonical: https://ai.info/news/aws-agentcore-marketplace-payments-billing
