AWS Hands AI Agents a Wallet: Amazon, Coinbase, and Stripe Bring Stablecoin Payments to Bedrock
A new AgentCore Payments service lets autonomous AI agents spend USDC to buy APIs, data feeds, and paywalled content on their own — pulling crypto rails directly into mainstream cloud infrastructure.
Amazon Web Services has just made one of the most concrete moves yet to give AI agents real spending power. On Thursday, the cloud giant rolled out Amazon Bedrock AgentCore Payments, a system built in partnership with Coinbase and Stripe that allows AI agents to pay for online services using stablecoins as they execute tasks.
The launch effectively bakes machine-native crypto micropayments into AWS’s enterprise AI stack, removing what has been a stubborn obstacle for developers: the need to wire up a custom billing setup for every paid API or data source an agent might want to touch.
What AgentCore Payments actually does
At a technical level, the new service is built on top of Coinbase’s x402 protocol, an open standard that revives the long-dormant HTTP 402 “Payment Required” status code and turns it into a real-time settlement signal between machines. AgentCore Payments combines that protocol with Coinbase’s wallet infrastructure and Privy, the wallet technology now owned by Stripe.
Preethi CN, who leads AgentCore at AWS, framed the design choice around flexibility. AgentCore was built to plug into any framework or protocol, and the team carried that same neutrality into the payments layer so developers do not have to follow every twist in the rapidly evolving payment-protocol landscape, or commit to one standard prematurely. Amazon plans to add support for additional protocols at the platform level over time, so existing agents do not need to be rebuilt as the ecosystem shifts.
Micropayments first, full commerce later
The first wave of AgentCore Payments is focused on small, programmatic transactions — think APIs, AI tools, MCP servers, and paywalled web content. These are the kinds of charges that traditional payment networks tend to handle poorly because the fees per transaction can outweigh the value being exchanged.
AWS has signaled that the road map goes much further than micro-billing. Future iterations are expected to support broader commercial behavior, including booking flights, reserving hotel rooms, and completing purchases on merchant platforms. In other words, the same plumbing being used today to let an agent pay a few cents for an API call is meant to eventually let it spend hundreds of dollars on a travel itinerary without a human in the loop.
Why USDC, Base, and Solana
Under the hood, transactions settle in around 200 milliseconds using USDC, with the action happening on Base, the Ethereum layer-2 network operated by Coinbase, and on Solana. Coinbase says the x402 protocol has already cleared more than 169 million machine-native payments, spread across roughly 590,000 buyers and 100,000 sellers — numbers that suggest the rails are no longer just a proof of concept.
Stablecoins are quietly becoming the default settlement layer for agentic systems for a few practical reasons. They settle in seconds, run twenty-four hours a day with no banking holidays, and can handle sub-cent transactions without the friction or fees that make traditional card networks unsuitable for that scale.
Part of a broader shift in how AI handles money
AWS is not moving on its own. The announcement lands in a week where the same idea — giving autonomous software the ability to transact in crypto — is showing up across the industry. MoonPay recently rolled out a debit Mastercard designed specifically for AI agents to spend stablecoins at online merchants. Earlier this week, Solana and Google Cloud unveiled their own service that lets agents pay for APIs and cloud resources on a per-request basis using stablecoins.
Not every researcher is convinced stablecoins are the only answer. A study from the Bitcoin Policy Institute earlier in the year found that, when placed in simulated economic scenarios, AI models tended to lean toward Bitcoin and stablecoins over traditional fiat currencies — a result that hints at a more pluralistic future for machine money rather than a single winner.
Solving the compliance problem, not just the technical one
One of the more revealing comments around the launch came from Coinbase’s Head of Infrastructure Growth, Brian Foster. He noted that enterprise customers have repeatedly told the company they want agents that can transact, but keep getting stuck at legal and compliance review. AgentCore Payments, in his framing, is meant to give AWS developers a managed path to financial autonomy for their agents that just works out of the box.
That positioning matters. The hardest part of giving AI agents money has rarely been the cryptography or the protocol design — it has been getting general counsel comfortable with the idea. By embedding stablecoin payments inside an AWS-managed service, with Coinbase and Stripe as named partners, Amazon is effectively offering enterprises a pre-baked answer to that internal review.
What this signals for the agent economy
The bigger picture is that the infrastructure for an agent-driven economy is being assembled in plain sight, piece by piece. Identity, memory, tool use, and orchestration have dominated the AI agent conversation so far. Payments — the ability for an agent to actually transact, not just plan — is now joining the list.
Whether stablecoins end up as the dominant rail, or share the road with cards, bank transfers, and other crypto assets, the direction of travel is clear. The default assumption in enterprise AI is shifting from agents that recommend an action to agents that execute one, and execution increasingly means moving money. With AWS, Coinbase, and Stripe lined up behind the same standard, that future just got considerably closer.
