Agents can pay. So why don’t they?

Transcribed by qwen3-asr-1.7b with a schedule prompt. Possible prompt artifacts, names heard more often here than in the unprompted Voice Memos transcript (Qwen vs Voice Memos): Allie Howe (1 vs 0), Keycard (1 vs 0), AAuth (1 vs 0), Monzo (1 vs 0), Omni Analytics (1 vs 0), Apify (3 vs 0). Check them against the audio before quoting.

00:00:00 — Apify's Actor Marketplace & Why Agent Payments Matter

Instant payments. A practical, pressing question. Now we have all the instant payment protocols, so the question is like, why is nobody is using them? Why don't the why don't the agents pay? Before I start, just a quick introduction. Apify, we're running a marketplace of tools for AI. We call these tools actors, about eight thousand of them, and the different industries like extraction, social media, size, consumer, software development, travel, different teams and general like automation. They help you automate your work on the back.

And these actors are created by a global community of creators, and so they build and publish their models to the users, to the customers, and just last month, we paid almost two million dollars to the creators for their actors, right? So it's like driving two hundred marketplaces. So obviously, by now you understand why we take great interest in agent payments because for us it's a standard way how we can sell our actors, our tools, and our writers to more people and agents. Right, and so obviously when it comes to new a way how we can integrate our system, our platform, our marketplace.

So, actually, last year at AI World Summit in San Francisco, I was the only one was talking about this topic about the topic of agent agreements and agent economic more broadly, because back then it wasn't really like you know well known, not many people talking about it, and I think like over the last year and a half, there's been great progress, right? So this year, on AI Engineering World Summit in San Francisco, there was a dedicated track from JPMorgan, and now you know, over the past year, I think like really everyone understands that agents, in order to complete ever longer tasks, they will actually need a budget, because you know at some point, they will ask you to do something, and so on. And if you want to do the other most, and really complete the bigger task, we might need to give them money, and there's been a huge progress over the past year and a half in this domain.

00:02:26 — The Explosion of Agentic Payment Standards

So, I think that doctor predicts that by 2028 this would be like a 15 trillion dollar economy, that 80 percent of B2B activity is in between companies will be done under all situations. I mean that might be much more seen, but anyway, a lot of players in the finance and finance in fintech jumped on this bandwagon because they understood they understand the size of this opportunity.

So we saw a lot of different tools and payment standards appear over the past, you know, year and a half. So it started L402 by Lightning Labs. It was made to those events or super early. And last year in April, Mastercard issued a — then May two thousand and five explored by Coinbase, it's probably the most popular agent development standard currently. In June last year, KYA Pay by Skyfire Berisa. We have Asian Business Protocol, AAuth by Google, the Agent Commerce Protocol (ACP) by OpenAI and Stripe in September last year.

In October, trusted agent protocol by Visa. January Universal Commerce Universal Commerce Protocol UCP by Google and Shopify. There's more agent decoders for Agent decoder trust protocol (ACTP) by Alipay. Machine-based protocol MVP by Stripe and Monzo. That's from earlier this year. And from April, Agentic Mobile Protocol AMP by Allie Howe. And last but not least, AgentPay payment open protocol API by Omni Analytics. Oh, there's more, OKX agent payment protocol by OKX.

So you can see this big, a little battleground for the future of Asian newcomers, because all these like players understand, okay, Asians are gonna pay at some point, and this company is actually going to, you know, a part of the of the of the future, and preferably to use their networks and their infrastructure to do that — oh, one more, Agent Pay for machines by Mastercard. Sorry, there's all of these, and it's really hard to keep track of all the news stories happening right now.

00:04:45 — Why Crypto Fits Machine Payments & Real-World Volume

And I think, like this whole industry commerce became like a huge, huge, well, big news for crypto people because, finally, there seems to be like a real use case they're really waiting for a long time, you know, for crypto, except for like buying physical substances and hiding money away from your spouse.

And I actually believe that. Like I'm not like a crypto person myself. I'm not pumping anything. I don't make crypto, but I think like crypto is actually very well positioned to when things like machine payments worse because, crypto, like if done well, is it doesn't have like single order, there's like no single company controlling the ecosystem, right? And properties that make it very useful for payments. So, for example, compare crypto bit like credit cards, PayPal, or PayPal micro payments, or you know bank debit.

It's much cheaper. Like you just can't like use differential payment methods for small everyday transactions. It just doesn't make economic sense, and also, the traditional payment methods they also chargebacks. Like everybody who's selling online understands that we have to count that, like some people just do surveys and then ask for money back, and it's just like you know, way of life for online businesses, but you can't do that in the in the world, because there's like infinite number of agents, and they were just like, you know, steal a little money effectively, or you know, they can do this for free. So on the other, there are some cardbacks. So only like refunds if you bought these are optional by the merchants.

So I think that really like Keycard is well suited for Asian companies, and it is like bright future for that, and actually, two oh sorry, two biggest players — two most popular technical standards currently explored by Coinbase and stripe MPP on stable pass — actually, like using crypto, right? And it's not surprising.

So let's have a look at the adoption of these two systems in real world. This is from the X-ray bluescan and NPP scan websites, which basically monitors transactions on blockchain, and you can see in the last thirty days, in the last 30 days, on this road too, there was just only like one million dollars. Nobody. And on MPP, only like one point five thousand dollars. I mean, it's basically nothing. It's like non-existent. So the question is like, why is everybody — why is nobody using it? Like it's been more than like year and a half. Now, it was the kind of thing, but he starts to look into it.

00:07:35 — The Agent Payment Funnel & API Key Competition

So let's try to answer to that question. So first, maybe there's a low supply services. Maybe there are just not too many things to buy on for agents, mostly. Well, I don't think that's the case because if you look and these actually payment directories, like there's like multiple services, you know, and I would say just like, our company and Amazon, for instance, in particular, are all there, and I would say we together make far more, you know, than million dollars per month from agents using for citizens, right? So people are using these tools, pay much more for that, but to EVITs, not to the most discovery and demand. So it must be the demand, right? So clearly, the agents just aren't interested. They aren't buying the services. So let's try to understand why.

Okay, and I think the best way to do that is to look at the engagement parents as a funnel. So it starts with like need. So first, you as author of the system or creator of the agent, internally, like okay, because otherwise, like we wouldn't allow it, and so on, and so forth. Then you like someone also set it up, and once it's once it's done, the agents need some way to discover the services and eventually buy them, right? I would say at every step of this funnel, we have some friction points that ultimately, like lead to the fact that like agents aren't buying at the end of the funnel.

So let's try to look at it more in more detail. So first, the need. Well, most people are building agent assistants now, and they give them access to services, but they do it through fixed accounts or like you know accounts with sign up. I don't know, like you have an engine, and if you need access to the data, you just create an account on Apify, and then like put the API key into your cloud, or you know to use MCP to connect the services, and then so and then you pay for the use of the services through your credit card, right? It's very straightforward. Everybody can do that.

So it's a little like that. Like authentication are competing with Visa. It's they are competing with API keys because it's just the more convenient way.

So, what is to happen for agents or people to relate decide, okay, I want my agent not to use only Pixel for services that I you know gave it before, but I want my agent to also discover new services as it goes, and buy those on the phone?

So, I think there's three prerequisites to that. First is better models. I mean, as the models are getting better, the agents can complete ever longer tasks. So mostly, you trust them more, and so on. So this is sort of about controversial things, and I think over time, the models like, if they don't have it already, will they ever trust also with money. I think like you can start like giving them I don't know five dollars per job, and then fifty, and if you know it works, like you can do five hundred. And I should say, people are in trust agents with access to their systems, even right access, quite often sometimes leads to disasters, or you know, like you know, unless well, I'm gonna leave your email as my first. So this is already progressing. Like models are getting better, and they will have more trust for people to do that, to British services. But I think big, another one coming to always on agents. Now we saw with like Meta News, you know, and I Dots and so on. I think the personal always on agents are becoming a real thing.

And if you want those agents to actually do something, like you will, you'll debatch it eventually. So I think we're getting to this stage as well.

And I think what will also help a lot is end of the era of open subsidies. Now, on $100 per month on Claude, you can buy I don't know like $14,000 worth of tokens if you box it out. This is crazy because then the agents have no incentives whatsoever to, you know, to look for existing services because they can build anything from scratch because it costs nothing. But I think at some point this will end, maybe with OpenAI not doing this IPO, and some of the Asians will be okay, maybe it's more accomplishable for me to buy a service rather than build it from scratch or figure out where to go get it done. So I think this is like a lot is already happening.

00:12:04 — Wallet Setup Friction & Native Platform Integration

And then the second step in a in a in a funnel is setup. Okay, I want to use it now. How do I do it? I think for a lot of people, like crypto, is still very cryptic. You know, there's like all different like lingo, there's like different words, and you know, terms and like I would say like most people really haven't tried to set up crypto for themselves. Maybe they use like for trading, but we're like if they're interpretable, you know, it's you know it's like you know do some extra work from them, and it's a little more difficult, you know, than, I don't know, using a credit card to pay for a service because everybody's credit card ready.

But I think this is this is also like, you know, eight hundred thousand users experience. I think the agent development tooling is getting better, like every month, really. And I think big unlock would come if these major agentic platforms, you know, like Claude or GPT or you know these personal agents, if they actually integrate payments directly itself, so basically you just like use your bank account credentials from Claude and say like, hey, you can use a few dollars to purchase services. I think this will be like a big unlock, and I think it's coming sooner than later. All right. So that the agent has a critical one, or the payment system set up at the provider.

00:13:19 — Service Discovery, Trust & Verified Connector Directories

So, the question is now: Where does everybody find the services they can trust? And trust is important because if you're buying a service from some from someone, you need to make sure, okay, that they will deliver the service they promise. They will do, they will deliver for the price they say they will, right? And if you don't distrust, you wouldn't trust anyone.

And like history played. How are how are people building this trust? You know, like first, like they start to transact with people they know. So they like exchange services, a little bit of money, that was always someone they knew, so they didn't really take a trust, and they will solve the good thing, or they will translate to someone, someone else, like the new process, like recommended. Right, so it's like I was like, you ask, oh, if you're a good barber, like a friend, you'd be like, okay, you can use this one. You know, then you can trust the person, and then it's all great.

And over time, you know, as the world like got bigger, was I going to use like a trusted like a middleman to tell you who is like a good seller and who's not right? So, for example, like different marketplaces or downtown squares, where you know, if there was some someone like selling in the town, you know, on the square, you could probably trust, okay, they're fair and you know we have like different sort of like reviews and you know different agencies for consumer protection and so on. Also, this is a — these elements that give trust to the to the equipment that you can buy services from strangers because there's like signals you can use.

But in the agency world, there is nothing like that. How can we know that about this agent or the service on some random URL that claims, oh, I have API four. I know, better data, interested? Well, it's kind of hard, but I think the same principle will apply. You know, so interlike network of trust, like oh, I know these services are good, and also this is the public news, or basically using trusted directories and marketplaces, and we can already see this happening. For example, Called as a as a directory of connectors, right? It's like sort of like verified connectors that you know that, actually, like, another people are very, very careful. It's hard to get there, you know. So they provide all this layer of trust for the entirety of these tools, and same for ChatGPT, they have apps, and Apify is in marketplace as well, so basically there is just a middleman that tells you, okay, this is the service you can trust to know, and you're going to get results. So I think this is already happening as well.

And the — again, like a big cloud would be like these, like big providers enable like payments onto the box. Suddenly, I think the whole thing Google Star. And last is Dubai, right, and traditional these agentic payments systems they only supported this super simple use case and it was like fixed price per API call, but in this world, this is not sufficient. You know, there's always so many things you can buy for fixed price, maybe like the weather forecast.

But to complete a real job, you actually need like paper use. I mean, like, for example, to launch some compute job or some agentic job that uses LLMs itself, right, and so we need like variable pricing, and thankfully, just earlier this year, Expedia too added this option for best settlement, and NPP are its sessions, and the goal of these payment schemes is to enable agents, you know, to handle also this like paper use billing. So let's have a quick look on the how explore different types of paper use because I think this is a topic, nice illustration, what the early adopters had to go through, you know, to make it work.

00:17:02 — Implementing Pay-Per-Use & Batch Settlement on x402

So just a quick look how XForward works. Like first, like you know, you send a GET request, you get like 402 error back, it's some it's some payment signature. Sorry, payment required header. You pay the signature, repeat request, it's some money. Let the server verifies with the facilitator, which might be Coinbase, for example, it's all good, and then you do work there. Well, we'll see. Then you settle the transaction. Coinbase is a facilitator, something is transaction to blockchain, and transaction is software confirmed. You get like settled response back from the from the facilitator, and then, you know, actually, the money has been transferred, and you can do the job, right?

So the risk is like if you do the job a little earlier, and then the money doesn't arrive, like you can lose money. So it depends on your service. Sometimes you just want to make sure that you charge before you actually do the service, if there's like non-trivial cost, maybe like you know some compute, and then you get payment response. So this is all good, but this is like really just for the simple scenarios.

But handling paper use was pretty much like a little more challenging, you know. So first in May, Experian launched Exa Payments scheme. This was the first scheme for the fixed expressing. And then they released promise that will release a two paper scheme basically, allowing like this kind of low costs. They released earlier this year, but unfortunately, it didn't really work exactly for our use case, I'll get to that very quickly, and then finally, there is Bestelman's two thousand twenty-six.

So, brilliant Atul Berman's key was that, even though like it allowed you to charge random costs, there was not nothing preventing the buyer, the agent, to spend those money wise, you know, because the transaction was only like settled on blockchain later. So basically, up to wasn't really useful for paper use case, which was a bit of a bummer.

So we work around it by using exact refunds, so that you don't be charged like I don't know, like ten dollars upfront, and then you know, I think it was two dollars. And once the board was done, we were handed the rest. But then again, that means that the agents, the seller, which you get, or the origin point, it's difficult sometimes. So and it has two-person transactions, so that means like more fees and so on. So this wasn't really ideal.

So finally, the Basic Life Insurance Scheme saw this issue. We added it just recently, I think last week. I don't have time to go to the detail, but it uses a pretty magic kind of like escrow the money on blockchain, and then like you can like charge it like you know for your payment transactions later, and...

Final remarks. So I think the independent funnel is really being ended right now, and soon it will really like start like a real volumes on the on the and real startup digital economy, and I think it's really good time now to figure how we can sell your services in this new world because the implementation might be a little more difficult, I think, and if you wait too long, if you wait too long, maybe, you know, as a friend.