Amazon Web Services (AWS) has made Amazon Bedrock AgentCore payments generally available, giving AI agents the ability to autonomously pay for APIs, web content, and other services through supported digital wallets, including Coinbase and Stripe Privy wallets.
AWS announced the general availability on August 18, following a preview launched in May with Coinbase and Stripe. The service is designed to let enterprises deploy AI agents that can make payments without requiring a human to approve each transaction. The crypto angle is significant because AgentCore supports USDC-funded wallets and machine-payment protocols such as x402 and MPP, giving AI agents infrastructure to pay for digital services programmatically.
AWS said the launch is designed to give enterprises agentic payment capabilities with security, guardrails, and observability.
How agent payments work
AgentCore payments connects AI agents with Coinbase and Stripe Privy wallets, allowing developers to provide payment capabilities through wallet credentials. End users can fund an agent’s wallet through traditional payment methods such as credit cards or USDC. They must also authorize the agent to spend funds on their behalf.
AWS said raw developer credentials are not exposed to the agent. Instead, credentials are stored in AgentCore Identity Secrets Manager, while short-lived tokens are used to direct wallet operations such as transaction signing.
For Coinbase, AWS has added a “Quick Create” option that lets developers provision credentials directly through the AgentCore console or CLI. Stripe Privy users must obtain their credentials through the Privy dashboard.
This structure separates the AI agent from the underlying wallet credentials, allowing the agent to initiate authorized payments without receiving unrestricted access to the user’s wallet infrastructure.
Stablecoins and machine payments
The system initially supported the x402 payment protocol during its preview phase. With the general release, AWS has added support for the Machine Payment Protocol (MPP), a standard co-authored by Stripe and Tempo.
AWS has also added the “upto” payment scheme for x402. Rather than requiring an agent to commit to a fixed price, the system can establish a spending ceiling and charge based on actual usage.
That model is particularly relevant to stablecoin payments because AI agents can make small, automated payments for digital services without requiring a human to manually approve every transaction.
Potential applications include AI inference, computing and usage-metered APIs, where the final cost depends on how much of a service an agent consumes. AWS said the feature can support “pay-per-inference” and dynamic pricing models, allowing merchants to charge based on the amount of a service actually consumed.
Spending limits built into transactions
Autonomous agents can behave unpredictably, particularly when they encounter unexpected responses or retries. AgentCore payments therefore places transactions inside payment sessions with two configurable limits: a maximum spending amount and an expiry time.
Before a payment is signed, AgentCore checks whether the transaction would exceed the session’s budget. Requests that breach the limit are rejected at the infrastructure layer.
AWS is also adding payment monitoring through AgentCore Observability. Developers can track payment audit trails, logs and metrics such as transaction success rates and average transaction values through Amazon CloudWatch and AgentCore dashboards.
The controls are important for crypto-based agent payments because giving an AI system access to programmable funds introduces a different risk profile from conventional human-initiated transactions.
From paywalled content to AI inference
AWS sees several potential applications for the system. Agents can use it to pay for web content that would otherwise be inaccessible behind a paywall. AWS said it is working with content delivery networks including Amazon CloudFront and Cloudflare, while browser automation company Anchor Browser has integrated AgentCore payments for paid web content.
Cloudflare Chief Strategy Officer Stephanie Cohen said its Monetization Gateway allows asset owners to “charge agents directly, per request and at a price they set using the open x402 protocol.”
Another emerging use case is pay-per-inference. AWS said SpreadX’s AI product Incarna uses AgentCore payments to pay for inference through BlockRun, a routing and payment layer that uses x402.
The system is also being used for agent-driven travel and research. Travala has integrated AgentCore payments into its Travel MCP servers, allowing users of agentic platforms such as Claude to book hotels conversationally. Elsa AI and Heurist AI are using the service to access pay-per-use financial research and advisory APIs.
BlockRun founder and CEO Vicky Fu said AgentCore payments provides “a managed service that makes it easy for agents to securely access and pay for BlockRun’s pay-per-use endpoints.”
A step toward an agentic economy
The launch reflects a broader shift in AI infrastructure: agents are moving beyond generating responses toward independently carrying out transactions.
For crypto and stablecoin markets, that creates another potential source of demand for digital-dollar payments. Small, automated transactions can be difficult to handle efficiently through traditional payment systems, while stablecoins can be used for digital payments across machine-to-machine services.
AWS’s approach, however, keeps the payment process within a controlled infrastructure layer, with spending limits, wallet authorization and transaction monitoring designed to reduce the risks associated with autonomous spending.
With AgentCore Payments now generally available, the next test will be whether developers adopt autonomous payments at scale and whether x402 and MPP gain broader use among AI service providers. If adoption accelerates, AI agents could become a new source of transaction demand for stablecoins such as USDC.
Also Read: Arthur Hayes Comes Out of Retirement to Launch New AI Token
