Editor's Note: This article offers a relatively calm builder's perspective. Over the past year, agent payments have become a hot narrative in the intersection of AI, payments, and crypto. Companies like Stripe, Visa, Coinbase, and Google are all making moves, with concepts like stablecoin micropayments, x402, machine-to-machine settlements, and agent commerce gaining traction. However, the author, after actually building products and engaging with merchants and developers, found that genuine demand hasn't emerged at scale.
The article deconstructs several typical scenarios: Agent shopping isn't better than traditional e-commerce for most product categories because users still need images, comparisons, and browsing. Machine API payments seem suited for stablecoin micropayments, but most developers already solve this via subscriptions, pre-paid credits, and existing billing systems. Payments between agents, while a long-term vision, remains in its early stages with a lack of real transaction volume.
Relatively speaking, agent finance is one of the few areas with existing demand. Funds, treasury teams, and DeFi users already pay for financial tools, and AI can bring tangible capability improvements like real-time monitoring and automatic portfolio rebalancing. However, this market also favors traditional institutions that already possess licenses, compliance infrastructure, and customer relationships.
The author's final assessment is: What the agent economy truly lacks isn't just a payment layer, but more complex coordination capabilities—how to get agents and humans to collaborate, verify task completion, and settle results. Payment is just one piece of the puzzle. For giants, early positioning is a defensive choice; but for startups, what truly matters is finding the market that exists right now.
The following is the original text:
For the past year, I've been building infrastructure for the Agent economy and have also spoken with teams at Stripe, Visa, Coinbase, Google, and dozens of startups working on Agent commerce. I mapped this space, launched a product, and tried to find a real market.
But the reality is: Genuine demand hasn't appeared yet. For startups wanting to enter this field, there are still many structural issues.
Stripe released 288 new products at its Sessions conference last month. Traffic to its Agent-related documentation is approaching 40% of all documentation reads. Its Agent Commerce marketplace has integrated with over 1000 merchants. However, at the Sessions venue, the number of Agents actually registering and completing transactions was only in the single digits.
Visa mentioned that its Agent tokens currently require a 3-to-9-month KYC approval process, and essentially require companies with annual revenue of at least $250 million to be eligible for access. Today, only companies on the level of Amazon and Walmart have the capability to close the identity verification loop.
Coinbase reportedly stated that by April, there were 69k active Agents and 165 million transactions on x402. But independent on-chain analysis shows the real daily transaction volume is about $17k, with roughly half of that being test transactions (CoinDesk, March 2026).
What We Learned Building shop.fast.xyz
Agent-to-Merchant, or Proxy Commerce
We built shop.fast.xyz to directly test proxy commerce. Real products, real merchants, real transactions.
But for most product categories, the current AI shopping experience is distinctly worse than traditional e-commerce. When buying clothes, electronics, or furniture, users want to see pictures, browse options, and compare side-by-side. A chatbot-style conversation is actually a step backward: you replace a rich visual interface with a string of text dialogue. Humans shop with their eyes first.
Agents performed well on the part we thought would be hardest. They can understand what the user wants and handle requests like "similar to this, but a bit cheaper" quite well. The model layer is effective. But it cannot replace the experience of "looking at ten items at once, then picking one." You can add product carousels and interactive displays to a chat interface, but at that point, you're essentially rebuilding an e-commerce frontend inside a chat window. For shopping scenarios requiring visual comparison, we haven't found a compelling answer for why a chat shell would be better than the original e-commerce interface.
We do see demand on the merchant side, but it's more defensive. Merchants want their stores to be queryable by Agents, not because many consumers are shopping via Agents today, but because they're worried they'll be left behind if Agents become a mainstream channel in the future. This is the so-called Agentic Engine Optimization opportunity, but it's currently a "nice to have," not a "must-have." Merchants are preparing in advance for a wave that hasn't arrived yet.
Where conversational commerce can genuinely improve the experience is for high-frequency, low-decision-cost purchases where users already know what they want. The clearest example is food ordering. The market is big enough, frequency high enough, decisions fast enough—like "help me order Pad Thai from the place I liked last time." In such scenarios, a conversational Agent might win. But the major delivery platforms don't have open APIs. The only path is computer use, letting the AI operate the app visually like a human. This process is slow, fragile, and the inference cost doesn't make sense for a $15 lunch.
Another opportunity is online stores so complex they're genuinely painful for users. Think stacked discounts, promo codes, loyalty points, and messy checkout flows. An Agent that understands "help me apply the coupon, use my points, find the cheapest shipping, and complete the checkout in my language" can indeed simplify today's broken shopping experience. This is especially important for elderly users, non-native speakers, and cross-regional shopping; or in very specific scenarios with extremely niche, complex needs.
But both these opportunities require massive B2C distribution power. You're competing with DoorDash, Amazon for the user entry point. Consumer-scale distribution capability is the strength of existing giants. The supply side for proxy commerce is ready, but the demand side is constrained by user experience and distribution channels. More infrastructure doesn't solve these two problems.
What We Learned from x402 and MPP
Agent-to-Web/API, or Machine Commerce
We spoke with dozens of developers about their real payment needs. The pattern was nearly identical: today's Agent API usage is essentially recurring consumption, like compute, inference, data sources. Developers already have subscriptions, API keys, linked accounts, and billing relationships with core providers.
The typical argument for stablecoin payments is: The effective minimum cost for card payments on Stripe is about 2.9% + $0.30, making sub-$1 API calls uneconomical. But at today's low transaction volumes, pre-paid credits solve the problem. Developers top up their accounts in advance, and the problem disappears.
The deeper issue is the supplier marketplace. Most large SaaS companies don't want to offer fractional-cent, piecemeal API access. Their business model is multi-year enterprise contracts. Companies reliant on large commitment revenue will resist new pricing models that bypass this.
Machine commerce is structurally a long-tail market. It serves small services, vertical data sources, independent developers, MCP servers, etc. Protocols like MPP and x402 are a great fit for this niche. But by definition, this is a market for users with professional needs; and developers have historically been among the most reluctant to pay.
When Stripe Projects launched, it integrated 32 service provider partners, including Vercel, Supabase, Cloudflare, Twilio, etc., covering most core services developers use to build and deploy software, all accessible via existing billing systems. The top of the developer tech stack is already well served. The opportunity for a new payment rail lies in everything beyond those top 30 providers: it's real, but naturally smaller than the market space implied by grand narratives.
The logic is the same for content access. Agents are already constantly scraping and summarizing articles, and publishers are pushing back. But when content monetization truly arrives at scale, it will likely come through CDN providers already sitting between publishers and the internet (like Cloudflare which launched AI audit tools), or through bulk licensing deals between publishers and AI labs. The infrastructure opportunity will flow to existing players with distribution power.
What We Learned from Agent-to-Agent Payments
Commerce between Agents is the long-term vision, but it remains almost entirely theoretical today. No one has run any meaningful transaction volume yet. The truly difficult parts are being tackled by various startups, including Agent discovery, trust establishment, term negotiation, and dispute resolution.
Once this transaction structure truly takes shape, it will look completely different from existing payment rails. Neither transacting party has a human identity; latency requirements are sub-second; transaction amounts can range from fractions of a cent to millions of dollars; and it can involve multi-party settlement, not the default bilateral buyer-seller model of existing rails. When it does happen, we believe it will explode with extreme speed and scale.
This is precisely the long-term bet for dedicated settlement infrastructure, and that bet is real. But a "real long-term bet" and a "current market" are not the same thing. We were also among those proclaiming this market would arrive for months, and built an entire infrastructure around it over the past few years, including our distributed network. Theoretically, it can scale to over 1 billion TPS, latency under 50ms, average consensus time of 10ms. But we have to return to where the market is now.
What We Learned from Agent Finance
Arguably, this is the only category with real existing demand. Customers already exist and are already paying. Fund managers, treasury teams, and DeFi users already spend money on financial tools today. Inserting AI into existing workflows is a natural product path.
Agent finance will also create entirely new behaviors. An Agent capable of autonomously monitoring and rebalancing hundreds of positions in real-time can operate in ways impossible to replicate manually. There's genuine capability enhancement here, not just automation.
The challenge is the competitive landscape. The finance industry is highly regulated and relationship-dependent. Incumbents have licenses, compliance infrastructure, and client relationships. Startups can enter in less regulated areas like DeFi, or find areas where incumbents move slowly or where AI can create new capabilities giants don't yet have. But overall, the competitive dynamics in this area favor incumbents more than the previous three categories, because adding AI on top of existing products and customers is far easier than starting with AI and then trying to add products and customers.
An Honest Summary
So, why are people still doing this? Two reasons.
The first is incentive alignment. Large companies have enough cash flow to bet on a future that may take years to materialize. For them, the cost of entering five years early is a rounding error; but the cost of being a year late could be catastrophic. So they have to do it.
The second is cognitive bias. When your business is payments, every problem looks like a payments problem. The Agent economy needs a payment layer, so people go build a payment layer.
But payments are just one part of a larger problem. The truly hard problem isn't moving money between Agents, but how to coordinate work between Agents and humans, how to verify if things are done, and how to settle results. Payment is just part of settlement. Settlement is just part of coordination. And coordination is the real prize.
Large-scale coordination will naturally generate demand for settlement mechanisms. Payments will become one instrument in that orchestration, not the entire symphony itself. The companies that truly solve coordination will end up incorporating payments, not the other way around.
Most existing giants are defensively building for a future of "mass machine transactions." For them, the timeline isn't critical because they have near-infinite runway.
But startups don't have that luxury. We have to find where the market really is right now. We can't wait forever for the wave to arrive.
A year of building has led us to an unexpected direction. There is activity there, growing fast and underserved. It exists outside the four categories we mapped.
