TL;DR — Key Takeaways
- RippleX has added support for the Machine Payments Protocol to its XRP Ledger AI Starter Kit, allowing AI agents to pay for services using XRP or RLUSD through a standardized payment flow.
- The update supports one-off payments as well as payment channels designed for high-volume micropayments, while adding spending controls and destination allowlists for agent transactions.
- The bigger story is interoperability: AI payments are starting to look like another API call, making standards, authorization controls and auditability more important than which individual token ultimately dominates.
AI agents are starting to do things that used to require a human with a credit card. They pull data from paid APIs, rent compute, and query specialized models mid-task. The trouble is that most payment systems still assume a person is clicking “checkout.” An agent that needs to pay ten different services mid-workflow doesn’t have time for account creation and a card form.
That gap is why a handful of machine-built payment protocols have shown up this year. Coinbase pushed x402, built on the old and mostly unused HTTP 402 “Payment Required” status code. Google has AP2. OpenAI and Stripe have ACP. And in March, Stripe and crypto investment firm Paradigm launched Tempo, a payments-focused blockchain, along with the Machine Payments Protocol (MPP), a standard for how an agent quotes, authorizes, and settles a payment without a human in the loop.
Ripple just decided to build for that standard instead of competing with it.
“Four competing standards are normal for an early market,” said Mitch Ashley, vice president and practice lead for CIO & technology buyers and software lifecycle engineering at The Futurum Group. “The ones that survive are the ones for which enough counterparties agree to speak.” By that measure, building to MPP rather than pushing its own rail is the more informative move Ripple could have made.
Stripe didn’t build Tempo and MPP to sell a new coin. It built them because merchants kept asking how an AI agent is supposed to pay for anything, and the answer, until recently, was “it can’t, not without a human approving each charge.” Tempo went live in March, backed by Visa, Mastercard, Deutsche Bank, Standard Chartered, Shopify, OpenAI, and Anthropic. A standard only works if both ends of a transaction agree to speak it.
RippleX, Ripple’s developer arm, released version 1.1 of its XRP Ledger AI Starter Kit this week, adding support for MPP alongside the Open Wallet Standard, a separate spec for managing wallets across multiple blockchains without exposing private keys. A developer building an AI agent on the XRP Ledger can now let that agent pay for services in XRP or RLUSD, Ripple’s dollar-pegged stablecoin, using the same MPP flow Stripe and Tempo already support elsewhere. Broader stablecoin support still awaits a separate ledger upgrade.
The mechanics are straightforward. A service responds to a request with an HTTP 402 status and a price. The agent authorizes payment. The service delivers what was asked for. Ripple’s kit supports one-off direct payments that clear in three to five seconds with sub-cent fees, and payment channels, where an agent pre-funds a channel once and then sends cryptographic vouchers for each charge, with the ledger recording only a single settlement transaction at the end. RippleX says the voucher signing can handle more than 100,000 payments per second, which matters once you picture an agent making dozens of small calls in a single task.
“Interoperability is not a feature a vendor adds later,” Ashley said. “Buyers should press every agent-payment vendor on which standards it settles against today, and treat a proprietary rail as a cost they will pay to unwind.” That’s a fair test to apply here too, not just to Ripple.
The kit also adds xrpl-up, a tool for testing payment flows locally, and automated trading on the XRP Ledger’s built-in decentralized exchange, so an agent holding XRP could swap for whatever asset a service actually wants.
Jazzi Cooper, RippleX’s head of product, framed the update as an extension of Ripple’s existing positioning rather than a new bet: “Our job is to make XRP and RLUSD first-class options wherever developers are building.”
That framing has to be measured against where the market actually is. Circle’s USDC is still the default currency for agent-to-agent payments, and none of the coverage of this release names a paying commercial customer running MPP transactions through XRP or RLUSD today. The starter kit is in beta, with no disclosed transaction volume. What stands out isn’t adoption numbers; it’s that Ripple chose to plug into someone else’s standard rather than push the XRP Ledger as its own separate agent-payment rail.
For platform teams, the more interesting part isn’t which token wins. It’s that payment infrastructure is becoming something agents call the same way they call any other API. “Once an agent can pay, the payment endpoint is one more thing it can reach, and spending is the least forgiving version of that authorization question,” Ashley said. “Spending limits and destination allowlists belong at the same layer as every other decision about what an agent may touch.” Ripple built both into the kit.
“Proof will pace adoption more than protocol choice will,” Ashley said. “Before an agent moves real money, finance and IT need one answer to what it spent, under whose authority, and how that is reconstructed at audit.” That’s the real bottleneck, and it applies whether the currency settling the transaction is XRP, RLUSD, or USDC. The competing standards will likely consolidate to one or two over the next couple of years, and the survivors tend to have the broadest institutional backing rather than the most elegant design. This release confirms that AI agents paying their own way are close enough to real that Stripe, Visa, Mastercard, two major banks, and now Ripple are all building for the same handshake.
