AWS has added preview payment features to Amazon Bedrock AgentCore, letting AI agents buy access to paid services using the x402 protocol. The company also announced a major cloud deal with U.S. Bancorp to move hundreds of internal workloads to AWS.
Amazon Web Services Inc. has added payment features to Amazon Bedrock AgentCore that let AI agents make purchases, the company said today. The capabilities are available in preview through AgentCore, AWS’ service for building and running AI agents.
Bedrock AgentCore is designed as a toolkit for companies that want to create AI agents and manage the parts they need to work together. AWS said the service also helps with related development tasks, including sharing agent components across teams.
This news matters because it shows AI agents are starting to get built-in ways to pay for services, not just ask humans for approval every time. A payment protocol is the set of rules used to complete a transaction, and x402 is AWS’ chosen method here. For non-experts, the key point is that companies are moving toward AI software that can both act and spend within limits.
If you are building self-hosted agents, treat payment as a first-class permission, not an afterthought. Add per-session spend caps, audit logs, and a separate wallet or funding source for each agent role so one compromised agent cannot drain a shared balance. If you want compatibility with paid APIs, start designing around HTTP-based payment flows like x402 rather than hard-coding one-off billing workarounds.
The new payment tools are meant to let agents pay for access to services and data without human intervention in every case. AWS said a financial analysis agent could use the feature to buy access to stock exchange datasets, while a coding agent could subscribe to a cloud-based development tool.
AWS said that, in the future, agents will be able to buy items such as flight tickets on behalf of users. For now, the company is focusing on service access and other transactions that fit the current preview release.
The payment system uses x402, a technology developed by Coinbase Global Inc. and based on HTTP status codes. HTTP is the web protocol used by browsers and servers to communicate, and status codes are the numerical responses that tell a client what happened during a request.
AWS pointed to the 404 error as the most familiar example. In the x402 flow, when an AI agent connects to a paid service, the service returns a 402 status code with instructions on how to pay for access.
Once the agent completes the payment, it receives proof of purchase. The agent can then refresh the webpage and access the paid content.
Bedrock AgentCore also supports payments through digital wallets. AWS said agents can use funds held in wallets and make payments in stablecoin or fiat currency. Stablecoins are digital tokens designed to hold a steady value, while fiat refers to government-issued money such as dollars or euros.
AWS said it has partnered with two wallet providers so far: Coinbase, which developed x402, and Privy, a subsidiary of Stripe Inc. The company also said developers can set spending limits for each AI agent session.
That session-level control is paired with observability features that let developers track an agent’s financial activity. AWS said it plans to add more payment features over time, including support for protocols beyond x402.
“There will soon be more AI agents transacting than humans, and they need money that’s built for the internet,” Brian Foster, head of infrastructure growth and strategy at Coinbase, said in a statement quoted by AWS.
AWS announced the payment features alongside another financial services deal. The company said it has signed a broad cloud agreement with U.S. Bancorp that it describes as “one of the largest and most comprehensive banking modernization initiatives.”
According to AWS, U.S. Bancorp plans to move hundreds of internal workloads to the cloud giant. Those applications cover payment processing, wealth management platforms and commercial banking tasks.
AWS also said U.S. Bancorp will use its AI services to automate customer support workflows. The moves add to AWS’ push to tie together cloud infrastructure, AI tools and payments for enterprise customers.
