On Ep. 8 of Agentic Commerce, Simon Taylor, Head of Market Development @ Tempo, and Bam Azizi, CEO & Founder @ Mesh are joined by Sam Ragsdale, Founder & CEO @ Merit Systems to discuss blockchain enabling massive scale for agentic transactions, Merit focusing on open agent commerce with USDC on Base and Solana and more!
On Ep. 8 of Agentic Commerce, Simon Taylor, Head of Market Development @ Tempo, and Bam Azizi, CEO & Founder @ Mesh are joined by Sam Ragsdale, Founder & CEO @ Merit Systems to discuss blockchain enabling massive scale for agentic transactions, Merit focusing on open agent commerce with USDC on Base and Solana and more!
Timestamps:
Tokenized is sponsored by Visa
A world leader in digital payments, Visa is bridging the gap between traditional financial institutions and innovative blockchain networks, helping players in the payments ecosystem navigate the ever-evolving world of tokenized fiat currencies with confidence and ease. Learn more at visa.com/crypto.
Tokenized is also presented by Mesh
As the first global crypto payments network, Mesh makes it possible for anyone — or any agent — to pay or get paid instantly, from any wallet, on any chain, anywhere in the world. Learn more at meshpay.com
***
We’d also like to remind you that the views or opinions of our contributors today are their own and do not necessarily reflect those of the companies they are representing. Nothing we say should be taken as tax, financial, investment or legal advice, do your own research!
Music by Henry McLean
Sam Ragsdale 0:00
Our view is that in on a five-year time horizon, 50% or more of commerce will be intermediated by agents. You will have some user-aligned agent that goes out and brokers with the internet on your behalf. And if this happens, like most of the internet as we know it doesn't make sense, or the stack for it doesn't make sense, like HTML websites probably don't make sense. That's not the most efficient way to communicate to an agent. The browser, as a general-purpose viewer of HTML websites, doesn't make sense. The search engine and the search engine paid for by ads doesn't make sense. And then display ads for your traditional way of monetizing the relationship doesn't make sense. So you throw all that away, and then you start with something new. Welcome to Tokenized. The show focuses on stablecoins and the institutional adoption of tokenized real-world assets. My name's Simon Taylor. I'm your host, author at FinTech Brain Food and head of Market Dev at Tempo. And welcome back to another episode of our Agentic Commerce series. And of course, because we're doing Agentic Commerce, we have to have the one and only Bamazi back in
Sy Taylor 1:00
the co-host seat, CEO and co-founder of Mesh, Bam. How are you doing?
Bam Azizi 1:04
Good, good. Thanks for another episode. I'm super excited.
Sy Taylor 1:08
Yeah, this this topic just seems to get hotter and hotter. And joining us today is Sam Ragsdale, who's founder and CEO of Merit Systems. How are you doing, Sam?
Sam Ragsdale 1:15
Very good. Excited to be here.
Sy Taylor 1:17
I'm excited to have you. Tell everybody what Merit is and what it does, and then we'll we'll kick off the show from that.
Sam Ragsdale 1:23
Yeah, sure. So so merit has done a few things in its history for the past eight months. We've been focused on agentic commerce or a particular variety that we call open agent commerce, which means permissionless listing on the merchant side. There are four products we have in market: x4 102 Scan and MPP Scan, which are the primary marketplaces where merchants can list, and then users or agents can go to discover x4 102 or MPP services. And then we have Agent Cash, which is a wallet that is an MCP server that you can put into an existing agent: Cloud Code, Codex, Noose, Hermes, Open Claw, etc. And then more recently, we have Poncho, which is a fully vertical agent hosted offering. Try Poncho.com that allows you to access anything in the x4 102 or MPP ecosystems.
Sy Taylor 2:09
Lovely stuff. And we'll get more into who's using that and how they're using it in a second. But before I jump in, I got to remind everybody watching and listening that views and opinions of contributors today may be their own and might not reflect those companies they represent. And please, don't take anything we say as tax, legal, or financial advice. This is for informational purposes only. And I've got to remind everybody that this series is, of course, made possible by Mesh. Thank you, Bam, for doing this and going down the agentic commas rabbit hole with us. All right, let's go to the background, Sam. Let's talk about your background. Like, how do you end up in agentic commerce on chain? Like, what's what's your path there look like?
Sam Ragsdale 2:48
Yeah, so I went to school at WashU where I met my co-founder Ryan. We've been doing projects together for about 12 years. Out of school, I worked at Google Ads as a lowly entry level software engineer. I spent two years there, I didn't expect it to be as important a part of my career. But the thing that you learn being at the ads giant that is Google is the old business model for the internet. What funded the free and open internet over the past 25 years, the majority of my my life was ads. Was that the any publisher could put content online, and then when consumers came and saw it, a third party, the advertiser, would would pay for the relationship, and Google would make sure that was all fair. In the process, Google became the biggest proponent of the free and open internet as we as we know it. The more people search, the more people click, the more people crawl. Google makes more money, and so they became a the business model for the internet. Turned into a $300 billion a year ads monopoly, not according to me. So spent two years there, and then went over to A 16 Z in 2021 when when crypto started taking off. I fell in love with with Uniswap v2 at the time. We were doing some Ryan and I were doing some arbitrage on it, doing some MVV, kind of like understanding the early innings of it pre DeFi summer when volumes really took off. The thing I really loved about Uniswap was it was a 250 line codebase, roughly the core of it, and it was on track to do a trillion dollars of transaction volume. And there was something that just broke in my brain when I saw that as an engineer. I didn't even really know what a smart contract was at the time. I hadn't been, I didn't hold a lot of speculative assets, but there was something really, really cool about that-that a 250 line code base could do that, and you know, didn't require anybody to run. It would run forever, and it was economic infrastructure for humanity. It was three parties: there was the the trader, the arbitrager that would keep the prices alive, and then the LPs that would provide capital, and that it organized an immense number of people just by putting this specific 250 line smart contract up. And so I fell in love with that idea of crypto, the idea that you could build sort of these money Lego blocks and really affect society in the process. Went over day 16Z Crypto, spent three years there as a. Technical infrastructure investor, things like L ones, L 2s bridges, perps, all all the technical stuff. The other half of my time I spent on our engineering team building open source projects and putting them out into the ecosystem, hopefully to push the whole space forward. The one I'm proudest of is is Jolt, which I built with a small team, world's fastest zero knowledge virtual machine. I won't get into the weeds on that. Very proud of that work. You can find it at GitHub.com/a16z/joult, and then left about a year and a half ago, two years ago, to start the company. At the time, we we started working on open source incentives. We spent about eight months, nine months on that until we decided to leave it behind. We found open source changed overnight with the advent of agents. When it became possible for everybody to contribute to open source, whether they knew what they were doing or not, open source kind of has crumbled under that weight. And we decided that somebody else had gotten to disrupting it before us, and then sort of went back to the drawing board on the things that we're really good at. We found that we're a really good application layer team. We happen to be really good at payments, and we happen to be really good at moving money on blockchains. And came to conviction that the next application for stablecoins beyond cross-border payments was going to be agentic payments. That agents natively needed a way to pay, and stablecoins are the simplest way to pay that's global by default and sub cent fixed fees, and so they made a ton of sense for agents, and then have really leaned in since then.
Bam Azizi 6:28
Yeah, I think that's kind of the the new revolution that's happening, right? So I was chatting with Paul at Morfo the other day, and he said like why people should trust you. Like that's the whole point. They don't need to trust me. You're the most audited code on the planet, and you have like only 650 lines of codes. Like you can't really go wrong. Or like when you're seeing hyperliquid stats that with like 12 employees, they were like generating 100 million dollar per employee. Like this is a revolution that was not possible without blockchain, and I'm like fascinated to see like how how revolutionary a piece of technology can be. Plus, if you add agentic and agentic commerce to it, that would be like basically what we have in terms of finance on the stoids, because we have like humans they're doing like two transactions a day in average. Agents, they can do 1000s of transactions. We have 5 billion human being connected to internet. We will have hundreds of billions of agents. Just imagine that the number of transaction that can happen, and the traditional rails cannot really handle that. They are not built. The nature of the the technology is not built for agentic commerce, so I'm very very bullish of the crossroad of like agentic boss boss blockchain.
Sam Ragsdale 7:49
Yeah, I think a lot of people, you know, they say why now for microtransactions they've been tried 10s of times before. You know, the the first microtransaction startup is as old as the internet, right? Like they were done at the same time, the common thing is like you know that the 402 status code existed when they're inventing the browser. When Mark Andreessen and Tim Berners-Lee and Co. were working on the browser, the 402 status code they thought there was going to be micro payments on the internet, and there's been a micro payment attempt every year since then for you know the past 27 years, and a lot of people focus on okay, like the infrastructure didn't exist. There was no way to do sub cent fixed fees at scale, and blockchains now allow that. But I think the other piece is the the part that you mentioned, which is the cognitive friction of microtransactions was too high. You know, if you as a human like had like a little notification in your browser every time you wanted to pay one cent for an article, and you had to click that every time, there's like a limited number of times you're going to click it before you get really really sick of it. What agents allow you to do though is you kind of like express your intent, your like budgeting goal in natural language, and then the agent will follow it, and it doesn't care. It'll it'll you know it'll focus on that 24/7 make a very good decision on your behalf as to whether or not you should be paying for this thing, and all of a sudden you can compose you know 500 different calls for one cent each from 500 different merchants and produce something that's that's much more valuable.
Sy Taylor 9:10
I think that's such a great point because the cognitive load and the user experience is so rarely considered in this conversation. I would wager that you're familiar with HTTP. I think it's 418, which is just I'm a teapot, and more that HTTP code has been called far more than 402 because engineers just love to do that thing as almost like a ping, right? It's probably been used a whole bunch more. Once you take the cognitive load out of it and you move it somewhere else, you're expressing something entirely different. But take me inside the new pain that exists for engineers, for teams. You're inside core code. You're inside codecs. You're using cursor. You need to buy some stuff. What What's the new pain look like? What's. The new friction, if it's not that cognitive load on the user,
Sam Ragsdale 10:04
yeah. So, so I think, well, I'll I'll do two different target customers. One is the engineer, everybody at my company, and then I I would say that the new up and coming vibe coder, which I actually think is a different category for engineers, because the time to execution of a given idea is so small now. Like you know, it used to take me like weeks to get one of these projects done. Like I have like a backlog of side projects, and I get to basically pick. Or you know, four years ago, I would pick three a year because they would take me four weeks each, something like that. And I only have a limited amount of extra time now with Claude Code, Codex, etc. You know, it gets done basically instantly. The bottleneck ends up being like, how many accounts can I sign up for, and like A/B test different API keys? And so I'll like hop between AWS and Vercel and Modal and a few others, and pull all these API keys and inject them into an environment. And I have a bunch of friends that pre x4 102 and all this stuff had like a a massive .env in their on their machine somewhere that has just every API key under the sun with a credit card pre attached, and they would just you know dump it into their clawed code or their codecs, and it worked fairly well for them because they knew what an API key was. They roughly trusted their agent, and they they didn't think it would it would run away with it, and that was actually a really good user experience. I think x4 102 NPP tools like Agent Cache, you have that one MCP server in all of your agents. For me, it's Cloud Code and Codex. I don't really use anything else, and then you have access to all of the hosting infrastructure, all the GPUs, all of the databases, all that stuff that you could ever need, and it makes it much faster to try a project. You know, I can I can spin up eight different prototypes in a day, and I I never leave Codex or Claude Code. Maybe it's to click out to the preview and test out my little web app. But I actually think the the audience that I'm more excited about is the up and coming. You call them an up and coming vibe coder, somebody that thinks about themselves as sort of like just below a developer, but they have the capabilities of a developer because they have the agency. They have cloud code. They have codecs, but I think it's even people that don't think of themselves as a developer. They're a they're a go to market person. They're a salesperson. But now they can they can hop into codex, cloud code, cloud coworks one of these emerging products, and they can say something like, you know, I want to find all of the farmers outside of Bozeman, Montana, that do north of $5 million a year of revenue, and they type that into a Claude Cowork. What happens behind the scenes is Claude Cowork writes a program, right? It's like, okay, I have these 15 APIs, like I have a local LLC search, and then I have Google Maps, and then I have White Pages, and then I have like some people enrichment search. And if I combine these all in this way and write, you know, the 600 line script, I actually can execute on this. And so it goes ahead, and you know, pays a few cents here and there, and maybe it says, okay, like I think there's another 300 of these. This is going to cost $4 Are you good with this? And the user says yes. It executes the program, and then it throws the programmer away and gives you the CSV, and and so this person doesn't think about themselves as a programmer. They don't know what an API is. They don't know what an API key is. They don't know how to integrate with a server. What HTTP is? They don't know any of this stuff. They started from a high level intent, which is I would like to find all of the farmers outside of Bozeman, Montana.
Sam Ragsdale 13:22
They put that into an agent that knows how to execute it. It pulls all the tools out of the shelf. It gets the the outcome done, and then it throws away, and the user is very happy with what they've done. And so, I don't know if you'd consider them a developer, but they're definitely a net new market for API consumption that did not exist before agents, and I could not have conceived of existing before agents.
Sy Taylor 13:45
Bam, do you know folks like that? And and this sort of liquid software, this single use software, are they going to choose stablecoins, or are they going to choose some of the payment method? And why might they choose stablecoins in in this particular case?
Bam Azizi 13:59
I think stablecoin versus fiat will be buried, right? So it would be like you would use it even without like knowing it, right? The the term called stablecoin sandwich exists for a reason. Like fiat comes in, converts to stablecoin, crosses border, and then gets converted to another local currency and gets off ramp. So I think from like what Sam mentioned, those high intent type of interaction with agents-they really don't care. Like they want the cheapest, fastest, best option for the payment, and stablecoin will will be able to power that. And the main reason is initially it's just yeah, I can pay with credit card or this and that, but when it comes down to like fraction of a cent, and you have to do that like 1000 times, like guess what? The traditional financial rails breaks that level. It's it's a quantum level of payment, and it just simply doesn't work. And then that's why stablecoin comes in, and then it's. Slowly, you feel like okay, like if I can use that for that, I can use it for $100 transaction because I don't need to wait two days for the other party to receive the fund on the other side of it. So slowly you feel like okay, like I can use it for everywhere. Why I'm just picking like micro transactions? I think that's where it starts, but will end to own entire transactions between agents to agents or human to agents,
Sy Taylor 15:24
I've seen a lot of folks launching like cards for agents lately. I know Crossmint did it, and seen a few companies that are launching just that and just giving them virtual cards. And it's one of the reasons when we were building MPP, I was so keen to support cards in MPP as a first class citizen, and you know, obviously, I know Kai pretty well, and we were able to to do some stuff there that benefits everybody in cards, not just Visa. And so, to your point, Bam, I do think these things start to coexist. And Sam, I was reflecting on what you were saying at the beginning of the show that you know you had a stint at Google. Google, arguably, the greatest monetizer of the attention economy there is, you know, they take that gap between you know you're unsure about what you want to you've bought the thing and they monetize that gap extremely well. An agent tends to show up with an intent fully loaded, like it already knows what it wants. Now it's a question of how efficiently can you route it? And so, what do you think about like the connection between user intent and outcome? How how's that evolving? And and how do you think about it in your agent cash wallet scenario? Like, how are you connecting these two? Because I know the card networks are spending a lot of time on intent and mandates and auto stability and all that sort of stuff. What what does it look like for
Sam Ragsdale 16:45
you? Yeah, so so for us, the main lever today is is search, which is that there's a tool in in the MCP server, the Agent Cache MCP server called Search. This will allow you to search for anything in the x4 102 or MPP ecosystems, and so you you say something like, "I would like to do enrichment, or nano banana image generation, or file hosting, and then we serve all of the results to you along with any telemetry we have on how often they've been called, so that your agent can pick which one it wants to use. It's very important to us that this is like user aligned, like that we do our we we don't serve advertising in this, right? That this is as good of a tool result as we possibly can, so that the end user's agent can choose the tool that they want to use. There's kind of an interesting, unique challenge here, which is particularly in agent cache, we don't actually have telemetry. We don't proxy all of the calls, like we don't intercept all of the requests, and so we don't actually know which services provided which results because we don't want to look through everybody's data, and so we have to do our best that we can to take a given natural language string and convert it to a set of results that provides the tool that people want, and and we do. In in some sense, we rely on the end user's agent to make a lot of the decisions. Right, the trajectory is based on your Claude context or your Hermes agent's context, and we just provide a set of tools, a router to you know all 40,000 tools in x4 102 and MPP.
Sy Taylor 18:20
Hmm. Super fascinating. Look, um, I'm going to pause us here where we hear from our sponsors, but we'll come back on this point in just a second. This episode, if it isn't already obvious, is sponsored by our friends at Visa, the leader in digital payments. Visa's tokenized asset platform, or VTap, uses smart contracts and cryptography to help banks bring fiat currencies on chain, whether it's launching a stablecoin or deposit tokens or something else. VTAP allows financial institutions to issue fiat-backed tokens, improving financial efficiency and enabling programmable finance. Check out the links in this episode's description to express your interest in VTap. This episode is brought to you by Mesh, the first global crypto payments network, connecting over 300 wallets, exchanges, and payments platforms. Mesh enables anyone to pay and get paid instantly anywhere in any asset, Mesh makes digital transactions seamless, secure, and universal, fueling the next era of agentic commerce. Learn more at meshpay.com/ai. Sam, just before you finishing, you were talking a little bit about agent cash. There, what's the response been from developers to what you built so far, what what are you hearing when you're speaking to them? Yeah,
Sam Ragsdale 19:46
so I think the the part that I'm most excited about is that there are a small and growing number of people that find it completely magical that they can put this one plugin into their codecs, their clod code, their Open Claw, and all of a sudden it. It's more capabilities, right? It's not this thing has a stablecoin wallet, and you know that can pay on Tempo and Solana and Base, or it can access x4 102 or MPP. It's just my agent became more effective as a result of putting this MCP server in place, right? Like now, all of a sudden, I'm able to host websites, and that's really cool. Or I can access all the frontier image and video gen, and that's really cool. Or I'm a market researcher that really likes to trade. Maybe I have the base MCP installed, and I I like buying things on Coinbase and formulating a a position on the equity or the crypto markets. And now I get to research all of the cross border trade data in order to do that. And with this one MCP server, I can access these 40,000 different APIs, and therefore my agent is much more effective. And I think that's that's really cool. One MCP, 40,000 different tools under the hood.
Bam Azizi 20:53
Do you guys support the card payments? As Simon mentioned, the agentic card helps you, or you don't care about the payment method.
Sam Ragsdale 21:02
We do not. So we support x4 102 on base and Solana and MPP on Tempo today. USDC across the board. We've talked quite a bit about supporting cards down the road. I think that the agent cash product particularly is focused on we call open agentic commerce. What we really care about there, the property that we think is really cool, and we wanted to lean into with that product and see if we could blow the market open, was open, which is that any merchant can list overnight. They can come online, you know, put up a server as easy as buying a domain in in 2000 and then we will be able to service them and allow any of the clients that have agent cash installed to to pay them directly, and so no no registry with any third party other than you know the domain service.
Sy Taylor 21:50
Yeah, part the um the 20 year payments veteran in me is like all chargebacks exist for a reason. Like this this this could be something that the market demands, but to your point, buying a domain in 2000 didn't have all of the downside risk yet. You just needed to make it easy because nobody's done it before. You can build in sort of protections to edge cases as as they start to evolve. So that makes complete sense. But I think the the dirty secret of like Explorer two and MPP is there's not a lot of volume yet. There's a magical experience if you've got it, but there's like almost more protocols than there are transactions. So, what are the what are the bits of kindling you're seeing, and and what do you think leads us to an inflection point?
Sam Ragsdale 22:34
Yeah. So, so I'm I'm not shy about saying this. I think I say this very aggressively. The big pickup in vol, you know, you'll see the numbers: 100 million, $250 million agent volume between x4 102 and MPP. The vast majority of this was in October, right after we launched x4 102 scan, and and the reason behind it was that many people launched tokens. They they had speculative activities that they were doing, and and the idea was, you know, let's make our volume look as high as humanly possible, and then we'll have a token associated with it, and people will be able to make lots of lots of money off of it. I don't think that's very productive towards creating new agentic capabilities. I think a lot of people had fun with it, but I think the real numbers are closer to 25 to $30,000 a month of volume, which is you know pocket change for like a Visa, Mastercard, Stripe, Coinbase, but but growing very steadily, and I think the the thing that's growing most steadily, and I'm most excited about, is merchants. So there, of course, the the the there was a massive uptick in October when there was that meme coin craze associated with x4 102 but since that leveled off, it's basically been linear growth since then, and it's starting to tick up in the past two months. And subjectively, I don't have an objective measure for this, but subjectively, the quality of the endpoints is increasing. There's there's things that I'm more likely to buy, and so I think my my old boss Chris Dixon has this famous chart of the the developer innovation price cycle, which is basically his view of of crypto and venture markets. Which is first, the prices will will spike, and then lots of buyers and developers and interests will come in, and then prices probably inevitably will dip. But the developers will stay around, and they will build for the next cycle. They'll build something awesome because now they're excited about what this technology could be. It probably didn't live up to it on the first price spike, but now they're around and they're going to build this future that they want to exist. And we very much see that happening, which is there was an initial spike associated with you know speculative activities that were not particularly exciting, but it did get the word out on the dream of what the agentic internet could look like. Like, what does it look like when every merchant has a headless offering that your clawed code, your user-aligned agent, can directly interact with? Lots of developers got nerd sniped by that. They're coming out. They're building API endpoints that will make the you know the next wave of the. In in six or 12 months, pretty awesome.
Sy Taylor 25:02
I think this is a great metaphor here, Bam, for the 2021 20 sort of there was the NFT craze and there was FTX and there was all of that sort of stuff in the Solana summer and stable summer came many years later, right? And and sort of, but what we got out of that was stable coins and payments infrastructure and embedded wallets and just like a whole bunch of builders came into the space back then. Bridge was founded back then, and and I think that's that's a really valid point. I mean, what do you think is being built now that's going to be really useful in in a few years' time?
Bam Azizi 25:38
Yeah, I think my prediction for the future is like everything will come on chain. So as a result of that, the crypto market cap goes from 4 trillion to 100 trillion. Not because Bitcoin price is going up, it's because all these real world assets are coming on chain. So you're already seeing tokenization of equities, tokenization of like all sort of assets, from real estate to debit to credit to lending, and and the benefit of that is like it is at the same time you're seeing this boom in AI. So the byproduct of these two trend would be amazing because if you think about it, I would I would compare it with another analogy that I have, which is like the movement from on-prem computation to cloud computation. In 2010, if someone asks you, do you think banks and all these big names would like bring their data on cloud? Like they should be crazy to just give up their security and then more than cloud. Like that that was the debate, and then fast forward to today, like majority of the computation is happening on the cloud. Most of the data is stored on the cloud. That firewall that existed, that boundary that existed, doesn't exist anymore. And I think the difference between cloud, the migration from on-prem to cloud, to off chain to on chain is you had to pay for the cloud for on chain it's free it's fast it's better so it's going it's inevitable and it's happening and the the beauty of it is the interoperability between basically different type of systems right so just imagine that your equity that is sitting on Robinhood can be now transferred to Fidelity, without like these two go and build proprietary integration with each other. If the asset stays on blockchain, everyone can interact with it, whether publicly or privately, without like knowing the proprietary knowledge of like how you interacted, how you put the asset on or store the asset. And I think that would be the beginning of a great era for AIs because they are they are faster than human being and they need the speed that those proprietary technologies cannot offer. But if everything is on blockchain, interoperability would basically maximize the speed of like interaction between agents, and I'm like super excited to see how this will unfold in in near future.
Sy Taylor 28:06
Definitely, I'm I'm interested in the let's let's go protocol for a second, Sam, and an honest assessment. How how you see that protocol shaking out? Obviously, I'll declare an interest. I like MPP, possibly, but but like but your honest assessment because x4 or two has Meta, it has narrative, it has a great name, it has you know been there for a while. It's adding new features all the time. There's MPP, but there's also like all of these other protocols from Google and Shopify and out in the universe, like how's all of this shaking out, or is it just not? Is it just like, hey, we could still not have HTTPS yet for a few years? We don't know how we're going to do any of this stuff.
Sam Ragsdale 28:54
Yeah. So the space that I see x4 102 and MPP playing in is that they play a very simple role in my mind, and this is how I explain it to merchants and clients that think about coming on board. It's for for agents. How do I pay? And for merchants, how do I check that the the the agent paid? The thing that's really nice about stablecoins is there's instant settlement. So checking that the agent paid is really simple, right? Like when somebody pays with a credit card, it's actually like ooh, like kind of extending credit to them, like I'm not sure it's going to settle in a month. What if there's a chargeback? Stablecoins are like cash, is what we've called it, agent cash. Either I have it or you have it. They're the closest equivalent to digital cash, and so x4 102 and MPP are well optimized for that. I am excited about some of the extensions, like you are in in x4 102 and MPP to support credit cards. I think then there's lots of other protocols that I think a lot of them are focused on agent to agent style stuff, where agents are selling directly to other agents. I haven't quite figured figured that out. It's very odd to me to think about an agent in a direct agent to agent negotiation when you negotiate. With another business or another person or another company, there's a fundamental limitation on it on like the respect for the other person, the number of rounds that you can do, and like the bandwidth of communication. If you know, just imagine models are free, like in the near future. If agents can just negotiate infinitely, then the game theory optimal thing to do would just be to state your preference up front, because otherwise you'll just go back and forth an infinite number of times. Like there's there's no cost in doing so, and so I think probably that will happen at some point in the future, but is not something I'm particularly focused on. And then I think the the payment layer is really just about how do I pay for this good or service as efficiently as possible? I think the other piece that we've been working on quite a bit, and we've worked with the Tempo team on this, and we've worked with the x4 102 Foundation on this as well, is the discovery layer, which is x4 102 and MPP will specify how to pay for the good or service once you're ready to check out. Say, but the version of the future we're really excited about is is sort of the headless internet. Like our our view is that in on a five year time horizon, 50% or more of commerce will be intermediated by agents. You will have some user aligned agent that goes out and brokers with the internet on your behalf. And if this happens, like most of the internet as we know it doesn't make sense, or the stack for it doesn't make sense. Like HTML websites probably don't make sense. That's not the most efficient way to communicate to an agent. The browser, as a general purpose viewer of HTML websites, doesn't make sense. The search engine and the search engine paid for by ads doesn't make sense. And then display ads for your traditional way of monetizing the relationship doesn't make sense. So you throw all that away, and then you start with something new. And sure, agents will pay either micro transactions or macro transactions for high intent, high ticket items. But then you also you need some way for them to discover all the merchants. You need some way for the merchants to say, you know, I sell Shopify stuff, or I sell enrichment, or I sell satellite data, or some other type of good or service that we can't dream up today, and so we've spent a lot of time figuring out that discovery layer, which is how, as a server, can I advertise my goods and services so an agent that is enabled with x4 102 or MPP can pay for that thing.
Sy Taylor 32:16
And walk me through what that discovery looks like today, because yeah, to your point, I might have a widget I want to sell, but how's anybody going to discover me? And I might want to buy your widget, but how do I know you exist?
Sam Ragsdale 32:29
Yeah. So, so today the the main thing we focused on is sort of the schema, and the way that we think about it is at the first level you have a an origin, so a domain name basically, and you can then fan out to all of the individual services that you sell. So maybe you have like five different pairs of shoes. Maybe you sell five different image and video generation models. And then each each one of those you have some parameters that you can pay for. Maybe in the first set of shoes you can pay for red, black, and gold. And the next you can pay for, and then there's sizes, and you can do that across all of them. And then at the very bottom of this, you also have some what we call like cross a resource context, which is what people traditionally think of as a skill. Here are the 15 endpoints you could possibly hit. Maybe for this one, you you first you request a let's say it's a flight checkout API. This is this is secret, but it's something we're currently working on. First, you you ask for a quote for flights to Charleston, South Carolina. Now you have a quote object. Now you want to maybe adjust the quote. You want to say I want a middle seat, and then you want to make another request saying actually I want to add another seat, and I I want to add a return flight, and then finally you want to check out and buy the thing, and then maybe later you want to you want to you want to cancel the flight and get your money back. Those are all different endpoints, and you have to compose them in a certain way in order to get the the result out of it. And so we have kind of those five different layers of context that, if you put it in a structured manner, the agent will know how and when to read it so that it can traverse your endpoint, traverse your merchant, and get the job done.
Sy Taylor 34:05
That makes complete sense. OTA or online travel is such a high risk, high fraud space. They definitely want to get paid sooner if they can. I'm not surprised. And also the fees involved in payments-they tend to be very sensitive as quite a low-margin business. I'm not surprised they're leaning in. So that reminds me as the the well-known endpoint in UCP. If you've looked at that, it's it's a very very similar model you're describing. So I wonder if these things all kind of come together. Bam, I'm interested in your thoughts on what you're seeing. You know, you have a pretty big global network where you're seeing this activity start to pick up, and and your thoughts on merit systems, and are you like me, somebody that has a tab open on both MPP scan and X402 scan permanently, or is is that not your vibe?
Bam Azizi 34:55
Yeah, I have many tabs open, but yes, 12 of us is scanning different. Things that's happening, but I think the first part of agentic commerce will happen on the macro side, whether we like it or not. I was like reading in the stat during Thanksgiving last year; more than 80% of American basically bought gifts based on the agent's recommendation, and the joke was like, "You trust agent more than your spouse. Like if I want to buy a gift, I just go to ChatGPT or Perplexity or any other agent, and then ask. And then just imagine that those agents can go and pay with PayPal or pay with Visa card and different type of merchants. That that will happen. That would be the initiation. That's where we are seeing the massive growth, and then slowly we'll see microtransactions, agents paying each other, or agents paying for the resources, whether it's API or image or whatever. And and that's that's where microtransaction comes to the picture. I strongly agree with Sam on in terms of like we need a new environment for agents. We need a new type of browsers and API model and protocols for agents to be able to communicate and interact with the with the internet. What we have like that you have to go to Shopify and click 20 different buttons and then check out that is that I think in five years nobody would do that. Like your agents would be capable of doing that. You don't even bother to type in like 16-digit codes and put your zip code or billing address. Those are all going to be automated and out of humans' way, but we are not ready today, as of now, for agents to be able to do those transactions. So the change is definitely necessary on the infrastructure, and I think what I am super interested to learn from, like, and also see from Sam and the industry in general is, are we going to consolidate those protocols, or it would be another stablecoin situation that every company has its own protocol? Well, they're
Sam Ragsdale 36:53
mostly ERC 20s, right? So at least Ethereum managed to consolidate there.
Bam Azizi 36:59
Yeah. So yeah, I'm I'm I'm curious to to see what happens in the market. I bet on consolidation. I hope the consolidation happens, but yeah, I know Simon. There there is NPP and exports too, but this
Sy Taylor 37:13
happened in the e-commerce space somewhat already, right? There was ACP and UCP, and those are starting to become one and the same thing, and I suspect ACP will get folded under UCP in the not too distant future. And the Google standards seem to have a bit of momentum behind them, with all of the PSP starting to adopt them. And that's to your point, Sam, about open. Like actually, you've got to take the community and the industry with you, but merchants are also ones that get left behind in a lot of these stories. It's PSP led, it's scheme led, it's issuer led. It's very rarely the merchant who gets to win here, and there's a new type of merchant coming online. So very excited to see what you guys are building at Merit. So, what's on the docket for you? What's the agenda look like? What what are you guys going to be doing next?
Sam Ragsdale 38:02
Yeah, maybe actually I'll answer that, but I would love to jump on the tail end of the previous comment, which is one of the things that I'm working on right now and working on with a few other people, a few other teams in the space. Is I think a fairly unfortunate state of affairs here would be that closed agent to commerce wins. That the only way that you're able to buy things through your agent is if your agent is whitelisted, and probably it's not actually your agent. It's not Sam's agent that's whitelisted. It's like Sam's Codex agent is whitelisted, or you know, OpenAI is whitelisted as a merchant that can buy from Shopify stores. I think that would be a fairly unfortunate state of affairs. I think the much more exciting one would be if every merchant sells to anybody over this protocol, the same way the web started. Right? It wasn't in order to render in Google Chrome, you have to make a partnership with Google with your website. Rather, your website stands up overnight, and now every browser-Firefox, Chrome, Internet Explorer, Netscape, Mosaic, whatever-is able to show you. And I think that's the future that's much more exciting. It's not bottlenecked by a few players. It allows for permissionless innovation. It allows for sort of the next generation of commerce that we can't possibly imagine today. And so that's the future that I, I, I very much am pushing for and why I lean for open agentic commerce. I think that the stablecoin standards naturally lean into that, which is it's really easy to pay anybody whether they registered or not. You do not need an intermediary. You do not need a central registry to do so. It's directly peer to peer between a user agent and a merchant agent. The other thing that I think, the other battle that's related that is going on right now is basically like we know people will have agents in the future. We know everybody will have an agent. Maybe you'll have a work agent. You'll make have a home agent. You'll have a consumer agent. I don't know how many there will be. And there's a battle basically for who owns it and like who it's aligned with. Today we're actually. A very interesting state, which is that the vast majority of agents are Gemini, Claude, and ChatGPT agents that users pay for directly, not monetized orthogonally. Which means that the the agents are user aligned. This this is true of Hermaz and Open Claw as well. That that agent is operating on your behalf, and and like Sam Altman has some quotes about like kind of the complexity of bringing ads in, which is that when you ask ChatGPT for a hotel recommendation, it'll try its very best to give you a good hotel recommendation, and it might not actually be the best one, but at least it tried. Like if you go to Amazon Rufus, for example, that agent is owned by Amazon and provided for free. What it's really provided for is the advertiser is paying for it again, right? Amazon now makes $17 billion a year on ads. It's a very significant line item for one of the largest companies in the world, and those ads are not user aligned, right? They're trying to sell you something whoever paid the highest for your impression, and so of course the agent will do that as well. And people might get used to a. The consumer might get used to a world where the agent is not user aligned. Actually, it's aligned with the merchant. And then we have some recent things coming out as well. Lots of people are leaning into local models. The local models are 99% cheaper and seems like only 20% less effective. If that trend continues, maybe everybody has a local model. And then Siri. It looks like we'll have a local model run by Apple that definitely is user aligned.
Sam Ragsdale 41:26
So two different battles: one on the merchant side, will merchant commerce be open, and then two, will the agents be user aligned or third party advertiser aligned? And I think both of them are fascinating. And I'm certainly rooting for the world where the commerce is open and the agents are user aligned, we'll see if that ends up happening. There's you know a lot of money in making sure the other way happens, but I think that there's a chance.
Sy Taylor 41:49
I love that you're out there fighting the good fight for that, and by pulling in that direction, maybe we end up a little bit closer to it. I do worry about the need to KYC agent users and KYB, the model providers, and with the geopolitics and a lot of that, it's just going to be hard to get away from. But user-aligned local models feels entirely possible. But then there's Mastodon and Blue Sky. Every every time you try and go like further into decentralized, you sort of end up in these like user experience problems, and so there's there's going to be a tension point, and we need somebody pulling in one direction. So, looping back then, so what are you going to be focused on over the next few weeks and months, and and what should we look forward to from you?
Sam Ragsdale 42:32
Yeah, so our main focus right now is Poncho. Try poncho.com Poncho is a vertical agent that has access to everything in x4 102 or MPP. One of the learnings we had from Agent Cache, which we're very grateful to our 1000s of Agent Cache users, but was that it was kind of hard to control the experience. It was hard for most people to install an MCP server. Not everybody is that sophisticated. Whereas Poncho, you can go to try poncho.com You can use the free tier. It's very generous, and you can immediately try Genta Commerce. One of the, as you pointed out earlier, we're talking about on the order of 20, $30,000 a month of volume. It's quite hard to get new merchants online. Some of the biggest companies in the world suggest that they host their APIs, maybe even create a new API to support agents. And so the real goal is to increase the the the size of the demand side. If I can get that number to 100 100 you know, a million dollars a month in volume, all of a sudden it's much much more compelling for new merchants to come online. And each incremental capability actually exponentially grows the capabilities of agents that have agentic commerce built in because it can compose together a bundle of 40,000 different resources that can provide various capabilities.
Sy Taylor 43:45
It's exciting times. I am looking forward to this more open commerce future. I want to see it exist, and I'm really big fan of what you're building at Merit System. So, if people want to find out more, if they want to go to tryponcho.com, but anything else they should be checking out, and where do they find? Where do they get hold of you, Sam?
Sam Ragsdale 44:04
Yeah, try poncho.com Easiest way to try out Agenta Commerce, or follow me on Twitter on x x.com/samrags_bam
Sy Taylor 44:13
How about you and everything at Mesh?
Bam Azizi 44:15
Yeah, you can follow Mesh Pay, and you can find me on x.com Bam, as easy mesh.
Sy Taylor 44:20
Perfect, and you'll find me at sy Taylor and all the socials, screaming into the void at fintechbrainfood.com, and also over at tempo.xyz, where we're cooking up some new white papers on all things tradfi. Thank you very much for watching and listening. It would really help us if you leave us a review and tell all of your friends to subscribe to, but we'll catch you next time. Bye for now.