AI agents made 1.4M payments on XRPL. Total fees: $280

The $XRP Ledger crossed 1.4 million AI-agent transactions this week, and Ripple joined Visa, Mastercard, and Google at the table writing the standard behind them. The milestone is real. So is the arithmetic underneath it: at a fifth of a cent per transaction, the entire agentic economy on XRPL has generated about $280 in fees, and the chain it is chasing has a hundred-times head start.

Every technology cycle produces a moment when a real trend and a modest number get announced in the same sentence, and the reader’s job is to hold both without letting either erase the other. The $XRP Ledger delivered this cycle’s cleanest example this week. The trend: autonomous AI agents, software that requests a service, receives a price, and pays for it with no human in the loop, are now transacting on public blockchains at meaningful frequency, under an open standard that Amazon, Google, Visa, and Mastercard have just formed a foundation to govern. The number: the $XRP Ledger’s share of that future crossed 1.4 million transactions, which, at the ledger’s fixed fee of roughly two-hundredths of a cent, works out to about $280 in total fees, on a network whose leading competitor has processed over a hundred million of the same payments with a year’s head start. Ripple’s engineering leadership frames the moment with a cloud-computing analogy, early days, obvious potential, standards still forming, and the analogy is fair, which is precisely why the honest piece about this milestone is neither the press release nor the dunk. It is the audit: what is actually being built, what the numbers actually measure, and what, if all of it works, actually accrues to whom.

The machinery: what x402 actually is

The protocol at the center of the story is elegant enough to explain in a paragraph, and its elegance is why the giants showed up.

When the web’s founders drafted HTTP in the 1990s, they reserved status code 402, Payment Required, for a payments layer the internet never built. Every online payment since has been a workaround: accounts, cards, subscriptions, API keys, invoices, all of them designed for humans with wallets and none of them usable by software that wants to buy one API call’s worth of data right now. x402, developed at Coinbase and contributed this month to a new Linux Foundation body, finally implements the dormant code. A service receiving a request from an unpaid client responds with 402 and a machine-readable quote: the price, the accepted asset, the receiving address. The requesting agent’s wallet signs and broadcasts the payment on a supported blockchain; the service verifies settlement and delivers. No account creation, no card on file, no human approval, no minimum viable subscription. Payment becomes a header, and commerce becomes something two pieces of software conclude in seconds.

The governance followed the code. The x402 Foundation launched on July 14 under the Linux Foundation with roughly 40 premier members, a list that reads like the payments establishment buying insurance on its own disruption: AWS, Google, Visa, Mastercard, Stripe, Circle, Coinbase, and, as of this month, Ripple. Membership is the context for everything Ripple has shipped around it: the XRPL AI Starter Kit released in June, packaging wallet integration, documentation servers, and payment tutorials for agent developers; the XRPL AI Hub launched by Ripple-backed t54.ai; support for agent payments in both $XRP and the $RLUSD stablecoin; and a slot among the thirty-plus launch partners of Mastercard’s own agent-payments network. The XRPL’s technical pitch for the workload is coherent: deterministic finality in three to five seconds, fees fixed at fractions of a cent, native escrow and multisignature support, and a built-in exchange, properties that suit high-frequency machine payments better than they ever suited the retail speculation the ledger mostly hosts. RippleX’s head of engineering, J. Ayo Akinyele, announced the million-transaction crossing with the early-cloud framing: “The potential was obvious, but the tooling and standards were still coming together.” As positioning, it is exactly right. As measurement, it invites the next section.

The audit: what 1.4 million transactions weighs

Take the milestone apart with the ledger’s own arithmetic, because the exercise clarifies what is and is not being claimed.

XRPL transaction fees are fixed near $0.0002. One million four hundred thousand agentic transactions therefore generated on the order of $280 in total network fees, a number that is not a gotcha but a measurement: it says the agentic activity on XRPL to date is, economically, a rounding error, and that transaction count on a chain where transactions cost nothing is a metric that measures enthusiasm and testing at least as much as commerce. At two-hundredths of a cent, a single developer’s integration test suite, a hackathon weekend, or an agent pinging a demo API in a loop produces six-figure transaction counts for the price of a coffee. Some unknowable share of the 1.4 million is exactly that, which the more careful voices in the ecosystem, including t54’s own framing of the milestone as showing capability, implicitly concede. The honest description is that XRPL has proven the pipes work, not that anything economically significant flows through them yet.

The comparative table sharpens the same point. Coinbase’s Base network has processed more than 119 million x402 payments; Solana roughly 35 million; both had approximately a year’s head start, and both settle the overwhelming majority of that volume in $USDC. Even the leader’s economics remain tiny, industry tallies put cumulative settled x402 volume in the tens of millions of dollars, an average well under a dollar per payment, which confirms the category is micropayments in fact as well as theory. But the ordering matters: XRPL’s 1.4 million against Base’s 119 million is a roughly hundred-fold gap in the category XRPL is now marketing as a strategic fit, and gaps of that shape, in developer-network businesses, historically widen rather than close, because agent frameworks integrate the chains where the other agents already are. The $XRP ecosystem has run this race before, shipping credible infrastructure into a category with an entrenched leader and discovering that technical fitness does not conjure developer gravity; the EVM sidechain’s first year, which this publication audited at $25,741 in total value locked, is the cautionary precedent nobody at the milestone party mentions.

And beneath both numbers sits the question this ecosystem can never quite escape, because it is the question: who earns what if this works? Agents transacting under x402 optimize for stable settlement, which is why $USDC dominates the category everywhere it exists, and on XRPL the natural settlement asset is $RLUSD, whose reserve income accrues to Ripple the company. The $XRP token’s role in the flow is gas, priced at two-hundredths of a cent by design, and collateral-adjacent plumbing, which means the milestone’s implicit promise, more agent activity equals more value through $XRP, runs directly into the fee math above: a billion agentic transactions a year, a seven-hundred-fold increase from today’s total, would generate roughly $200,000 in annual $XRP fee burn. The value-accrual gap between network success and token performance, the gap this publication has documented across payments, custody, and DeFi, arrives in the AI era fully intact. Machine commerce may be enormous. XRPL may even win a real share. The token’s claim on that outcome remains what it has always been: a thesis in search of a mechanism.

The case that the position is still right

Having weighed the milestone honestly, weigh the strategy the same way, because the audit cuts against the hype without cutting against the play.

Standards tables are cheap options on large futures. Ripple’s premier membership costs it engineering attention and puts $XRP and $RLUSD inside the specification process of a payment standard that AWS, Google, Visa, and Mastercard consider worth governing, which is not a marketing decision on their part; the agent-payments category is the rare crypto use case that the traditional payments industry believes in enough to pre-organize around. If machine-to-machine commerce becomes a fraction of what its backers project, the chains and assets wired into the standard from the beginning inherit distribution no retrofit can buy, and the Mastercard launch-partner slot is exactly that wiring. The early-cloud analogy earns its keep here: AWS’s revenue in 2008 was a rounding error too, and the companies that dismissed it on contemporary arithmetic were measuring the wrong thing.

The technical fit argument is also better than the ecosystem’s average claim of this genre. Agent payments genuinely want what XRPL genuinely has: deterministic sub-five-second finality, fees that never spike, native escrow for conditional payments, and an architecture that has processed payments, only payments, for a decade without an outage that mattered. The chains currently leading the category are general-purpose platforms on which payments compete with everything else for blockspace; a specialized settlement layer is a coherent bet on how the category matures, particularly for the enterprise and financial-institution agents Ripple’s distribution actually reaches, as opposed to the consumer-crypto agents Base inherits from Coinbase. And $RLUSD’s presence in the standard is unambiguously valuable for Ripple’s stablecoin strategy, whatever it does for the token: every x402 flow $RLUSD settles is float, and float is the business.

The bear case concedes all of this and returns to the ledger’s oldest pattern: infrastructure fitness without developer gravity, milestones denominated in counts rather than dollars, and value accruing to the company faster than to the asset. Both cases are live. The difference between them will not be argued into resolution; it will be measured, which is what the final section is for.

The stablecoin sitting in the middle

One participant in this story holds a materially different position from all the others, and the analysis owes it a section: $RLUSD, which enters the agent-payments race with none of $XRP’s value-accrual problem and all of Ripple’s distribution behind it.

The economics of a stablecoin in machine commerce are the economics every issuer already understands, at higher frequency. Each $RLUSD that settles agent payments is float, reserves earning treasury yield for the issuer, and agentic flows have a property consumer flows lack: balances that never sleep. A human cardholder’s stablecoins sit idle between purchases; an agent’s working balance turns over continuously, but the aggregate float across a fleet of funded agents is persistent, programmatic, and grows with the category mechanically. If machine payments become a fraction of what the foundation’s membership implies, the stablecoins wired into the standard become the category’s silent tax collectors, and the fight for that position is already visible in the data: $USDC’s dominance of Base and Solana x402 volume is Circle collecting the early category almost uncontested. $RLUSD’s presence in the XRPL implementation, and in whatever flows the Mastercard partnership eventually routes, is Ripple’s bid for a share, and it is a better bid than the transaction counts suggest, because the enterprise agents Ripple’s institutional relationships reach will care about exactly the things $RLUSD was chartered to offer: a regulated issuer, bank-grade reserves, and a compliance posture that a corporate treasury can sign off on.

Which sharpens, not softens, the token question this piece keeps returning to. The clearer $RLUSD’s path in agent payments becomes, the more precisely the ecosystem’s value routing resolves: the category’s fees go to nearly nothing by design, the float goes to Ripple, and the $XRP token’s participation is the $0.0002 toll. There is one construction under which the token does capture something, $XRP as the bridge and liquidity asset when agents transact across currencies, using the ledger’s native exchange, which is the on-ledger version of the company’s oldest thesis, and it carries the oldest caveat: it requires agents to hold and route through a volatile asset when a stable one is available, a behavior no current x402 flow exhibits anywhere. Watching whether it ever emerges, in the cross-currency settlement data the ledger makes public, is the cleanest token-relevant observable this whole story offers. Absent it, the honest summary of the agent era for the two assets is uncomfortable and simple: the milestone is XRPL’s, the business is $RLUSD’s, and the token is, once again, the venue, not the beneficiary.

What to watch

Settled volume, not transaction count. The category’s honest metric is dollars settled through x402 flows on XRPL, a number nobody currently headlines precisely because it is small. When it appears, in t54’s reporting, foundation dashboards, or Ripple’s disclosures, it converts this story from enthusiasm-measurement to commerce-measurement. Until it appears, transaction counts should be read as what they are.

The settlement-asset split. Watch what share of XRPL agentic payments settle in $RLUSD versus $XRP, and what share of cross-chain x402 volume $RLUSD captures against $USDC’s incumbency. The first ratio prices the token’s role in its own ecosystem’s newest story; the second prices Ripple’s stablecoin against the category leader on neutral ground.

A commercial workload with a name. The milestone that would actually move this story is one identifiable production deployment, an enterprise paying real money for real services through XRPL agent rails, versus the anonymous aggregate counts. Mastercard’s network going live with Ripple in the loop is the likeliest venue. One named workload outweighs the next ten million test transactions.

The gap’s direction. Base at 119 million and growing; XRPL at 1.4 million and growing. The ratio between their growth rates over the next two quarters answers the developer-gravity question empirically, and it is the same question the EVM sidechain’s first year answered badly. Watch whether this category rhymes.

The 402 status code waited thirty years for the internet to need it, which is a useful reminder that infrastructure stories run on timelines that make any single milestone nearly meaningless. The $XRP Ledger’s 1.4 million agent transactions prove the machinery works and prove nothing about who wins, the $280 in fees prices today’s reality without pricing the future, and the foundation seat is a rational option on an outcome no one can yet measure. The audit’s conclusion is not that the story is false. It is that the story is, so far, exactly $280 large, and that everyone quoting the transaction count owes the fee line alongside it.

Disclaimer: This article is for information and educational purposes only and does not constitute financial or investment advice. It describes early-stage technology adoption whose metrics are incomplete and fast-changing, and comparisons rely on figures reported by third parties. Nothing here is a recommendation to buy, sell, or hold any asset. Always do your own research. Information is accurate as of July 23, 2026.

Frequently Asked Questions