Agents Are the New Websites: MoonPay's Tony Plasencia on the CLI Bet and Agent Commerce

Transcript

Gabe (00:00) Hey everyone, it's Gabe from Shoal Research and thanks for tuning in again into the Shoal Signal Show. And I'm here today with Tony, who's an expert on agents and payments. and all things on chain when it comes to CLI, MCPs, all these things, and we're gonna get into it in a minute. But yeah, I'm pleased to introduce Tony, and Tony, could you just give a little bit of background about your experience post-Web 2 where you were at Thumbtack and at Uber working on Agents, starting with Griffin, and then kind of your transition to MoonPay.

Tony @ Moonpay (00:33) Yeah, for sure. Well, thanks for having me, Gabe. Started in Solana about five years ago, building Underdog Protocol, which was a digital assets API and worked with Art Basel, Solana Foundation, and Solana Mobile, was a Solana hackathon winner, grand champion, ⁓ and scaled that out, ⁓ then transitioned into Griffin, the whole idea of it. behind our relationship, my co-founders and I is ⁓ making crypto very easy to use. And we realized LLMs ⁓ were the easiest way to do so. ⁓ So we built Griffin, scaled that out to close to 80,000 users, ⁓ about seven, eight figures worth of revenue and volume and joined MoonPay not too long ago to work everything AI and agents.

Gabe (01:25) Nice, nice. And at MoonPay right now, you're working on a lot of the piping, right? So the CLI stuff, skills, MCP, this cutting edge distribution for agents specifically. Can you share your experience of ⁓ Griffin, which was much of a chat interface, right? Where the inference and the LLM was hosted within the application to this potentially new experience where the inference is hosted actually on the user's end. and leveraging your tools. So yeah, if you can kind of give your experience and how you're kind of seeing this transition from almost like owned chat inference interfaces versus the tools that, know, LLMs can reason on.

Tony @ Moonpay (02:09) Yeah, for sure. That's a good question. ⁓ So at MoonPay, focusing specifically on the CLI and open wallet standard, ⁓ and what we noticed with Griffin and in the back end, ⁓ a lot of what was happening was simply tool calls. ⁓ when a user would come in with some sort of intent and say, hey, I want to create this cron job ⁓ that DCAs into specific meme coin, or hey, I want to use Moby to research and then place a trade through my Robinhood ⁓ crypto or through Jupiter. All that was happening at the backend was simply tool calls. ⁓ I think there's two parts to this. Number one is the owned interface, the chat interface, quickly got very ⁓ crowded. ⁓ And it was really difficult. It's classic consumer problem where it's difficult to ⁓ retain individuals. ⁓ So I think as we joined MoonPay, there was a lot of different initiatives and one large initiative to recreate what we've done. And ultimately where we settled was coming from an approach of, hey, everything that's happening already in the background is simply a tool call. Where we're seeing a lot of traction with AI specific tooling and LLMs is on the developer side. So instead of trying to recreate a consumer experience with which we may or may not be we as users completely ready for. Let's create something that's developer and prosumer focused. As you saw with like Claude code and with the codex, ⁓ the wave is saying those that really want to use LLMs are willing to spend times within their terminals. And that's when we decided to launch the MoonPay CLI. So with the MoonPay CLI, you get around 40 different skills, MoonPay and community owned. And you're able to within your terminal on ramp swap. do research with prediction markets, place trades on prediction markets, ⁓ spin up a virtual account, and then a lot more. So that's how we approached it. And I think ⁓ in the future, as we as builders figure out what the right skills and workflows and strategies are around tool calls or skills, ⁓ consumer applications will become a lot more prominent.

Gabe (04:51) Interesting. And I guess, so before the way that your product with the Griffin was set up was you're kind of like the orchestrator and you piped in all the tools yourself where now there's kind of this transition towards being a spoke in the orchestration stack, which might be like a coworker or a Claude. At what point did you realize this and...

Tony @ Moonpay (05:02) That's right.

Gabe (05:15) Yeah, I guess like start start with that like at what point or like what was the signal or like what was the kind of the ecosystem rumblings that that made you realize that the stack actually is a bit separated in that sense

Tony @ Moonpay (05:28) Yeah. So there is, I think, two parts to Griffin. Art one was the consumer ⁓ portion where anyone can go in griffin.com and start chatting with the experience. On the second end was actually the create your agent portion. So we allowed prosumers and enterprises to go in and create their own agents, turn the APIs into tools and feed that up into the into the consumer interface. ⁓ As we started to see more inbound from partners, we realized A, we can't do this at scale for individuals. ⁓ And B, a lot of enterprises are actually capable of doing this if they put a sole engineer or AI curious individual on top of it. ⁓ And so as we joined MoonPay, that learning of, hey, All of these are just tool calls. Hey, a lot of people are already a lot of AI, LLM native people are already using cloud code. ⁓ Let's go ahead and say, combine the LLM and the terminal with valuable skills that allow agents to move money. And let's see what comes out of it. And I think that's how we've approached building the CLI, ⁓ which is taking learnings from having enterprises and partners build their own agents on graphene.

Gabe (07:01) it. Okay, so you kind of had this, this stack or this this customer set of people who wanted to integrate their tools. And, and that was kind of like this, this process of, okay, we can't necessarily support all these specific integrations. And is that a good kind of understanding of that?

Tony @ Moonpay (07:10) That's right. Yeah, exactly. It's way to scale out what enterprises and prosumers wanted to build themselves within the interface that's most capable of doing it, which is a CLI or a terminal.

Gabe (07:37) And now you're building what would be the connector to it would what would be a service provider and it's kind of this transition from Okay, instead of trying to solve everything and kind of build this system where people will integrate There's like a standard, know the CLI the MCP connectors into these orchestration agents were typically they're like Claude codex, etc and so with that like where do you see the First of all, like the orchestrators, seems like they're almost being chosen, right? It seems like it's going to be open AI and anthropic. Do you see other environments where people will potentially use orchestrators outside of like those two frontier companies?

Tony @ Moonpay (08:20) Yeah, so I think I'll touch on the point ⁓ before, which is one of the big learnings that we had was really what ends up happening for a lot of these agentic products is a two-sided marketplace, where the agent is the supply side and the user is the demand side. And ultimately what you're trying to do is combine them so there's money moving throughout the marketplace. And with a lot of marketplaces, People usually say figure out the demand. I take the approach of like figuring out the supply to begin with. And the supply that we had to figure out was like agent, was agent liquidity. It's like, how do you find, and I've said this for a long time, is like, how do you find valuable enough agents, in this case now valuable enough skills that allow or bring in individuals to use those tools, use those agents, use those skills. And I think that's the problem that we're still, trying to face now and have done, I think, pretty well at MoonPay. In terms of what other orchestrators, I think it's gonna be the big three. It's gonna be OpenAI, it's gonna be Claude, it's gonna be Perplexity. Each one of them, I think, has their own niche. Perplexity, think, is really, really specialized in finance. So I can imagine a world where, like, Perplexity becomes the hub for... individualized finance agents. ⁓ Matter of fact, like I connected using a perplexity computer, I connected ⁓ MoonPay agents, the MoonPay CLI to some of the finance research that they've done. And I've been able to get like really in-depth prediction market ⁓ analysis or insight. think like Claude is very generalized and allows you to build a lot of these different ⁓ applications yourself. And then OpenAI I think is... ⁓ is in between, it's really like the best at everything, if you will. So I think those are gonna be the big three.

Gabe (10:22) Okay, interesting. Yeah, so you're seeing perplexity as like a specialized orchestrator essentially. Okay, verticalized, interesting.

Tony @ Moonpay (10:28) Yeah, exactly.

Gabe (10:31) Okay. And then for the piece about the kind of like the connectors and the hubs ⁓ for the orchestrator themselves, what are the use cases? And like we're trying to aim for like, you know, real use cases where customers are doing some type of process that is like legitimate, right? And not like, you know, watch trading or these points, you know, trying to accumulate points by using the CLR or something like that. But I'm curious on your end. Yeah. Like what are these use cases that you're seeing even if they're low volume, but almost like the high signal value driven use cases.

Tony @ Moonpay (11:09) Yeah, I think ⁓ number one is prediction market research to trade execution. These LLMs have access to like large troves of data. And I know some of the stuff you've done with Shoal, ⁓ like kind of proves that right? Where like it has this like, as this proprietary data has all these curated information that LLMs can use to make decisions with. think that's number one. Number two is combining a lot of this different data, whether it's like a Masari or BlockWorks or block mates or whatever it may be and use that for token trading. So there's a strategy that we have that's called, it's like VC Intel to funding where you're essentially able, an LLM is essentially pulling information from a sorry, seeing what categories of crypto VCs are investing in and is able to curate a basket of tokens for you. then ⁓ point you in the right direction to allocate capital. And then I think number three is like, commerce. ⁓ MoonPay has or acquired Solana Pay, which is now MoonPay commerce. So we have access to around 6,000 real world merchants within Shopify. And we're seeing people start to spend their money via the MoonPay on ramp being the buy widget or the virtual accounts. And they're doing that directly into Shopify. So whether that's like a Skylark, which is Justin Bieber's brand, whether it's like a Pudgy Penguins, or whether it's a smaller ⁓ boutique. sort of Shopify store. Those are the three interesting use cases and most use cases that we're seeing.

Gabe (12:43) Okay, interesting. Yeah, and that's a really good segue because I saw the UCP that was launched by Shopify as well and it kind of made me think of, we're gonna have like these headless merchants essentially where it's almost like a ghost kitchen and...

Tony @ Moonpay (12:50) Yeah. For sure. Ayayaya

Gabe (13:00) you know, especially coming from Uber, we're both at Uber for some time. The ghost kitchen thing was like the juggernaut that everyone was like, okay, it's thing gonna take over. And I think to some extent it did, but not as big as we thought. But you know, if you're looking for some toothpaste or something and you kind of delegate it to your agent, your agent can kind of traverse for good pricing and quality. ⁓

Tony @ Moonpay (13:02) Yeah. Yeah.

Gabe (13:22) That seems like an open landscape that a lot of people aren't really thinking about this headless merchant is a headless stores. I'm curious for you, like, are you guys thinking about this and how are you guys thinking about it? I guess the answer is yes, but more so how like what are the what are the second thoughts, the second and third thoughts and how deep are you? Have you been going on this?

Tony @ Moonpay (13:28) Yeah, for sure. Yeah, yeah, I think there's maybe like two or three thoughts around that. Number one is, yeah, I think like number one is agents deserve virtual cards ⁓ to have access to, or to be able to be backwards compatible with ⁓ how the internet is set up. So I think that's number one. ⁓ I think like number two is, Agents are the new websites. So instead of having to create your whole new website, which is fairly easy using Shopify, there's going to be those like large aggregators of stores or merchants. And all you have to do is work with an agent and make sure that your catalog is agent ready. So whenever someone goes in and types in like, hey, I'm looking for coffee mugs, your catalog comes up one, two, three, four, or five. I think that's the second piece. ⁓ I think the third piece is we're having really early conversations with some merchants. I can't really name who it is, but it's more ⁓ dining focused ⁓ where a lot of people are starting to think about how do I make, me as a merchant, how do I make sure that I'm agent ready? And it's going to be a collaborative effort where I think like, hey, ⁓ maybe I do need something that's backwards compatible with how the internet and the infrastructure works. Hey, like I do need to make sure that I'm agent ready and then ⁓ see it's like, hey, I actually have to have a good brainstorming partner to figure out how to do that. And I think with those three pieces together, we're gonna start seeing a lot more headless merchants and. think ultimately what Shopify released ⁓ is more on the merchant side, but that's a step in the right direction of saying like, hey, we wanna help make merchants agent ready. And we want to give merchants familiarity with agents and how they can use them day to day to run their business.

Gabe (15:41) Yeah. Yeah, interesting. The thing that comes up that I think through, like if I'm on Amazon, right, and I'm trying to buy a mug, let's use a mug example, I feel like a lot of the inputs into the mug for me are like price, you know, maybe I get tricked on some type of markup type of marketing strategy, right? Some discount strategy maybe influences my decision. The color of the mug probably like what the packaging looks like probably influences my decision. In the world of ⁓ agents, a lot of these tricks don't work anymore in terms of distribution, in terms of marketing. So how are you seeing the discoverability ⁓ for these agents and for these marketplaces? Do you have any ideas what the heuristics are that might be appealing to agents? Or is that just coming from the memory of the agent knowing the person?

Tony @ Moonpay (16:32) Totally. That's a good question. I don't think I've spent a lot of time thinking about it, but there's probably two pieces to it that come to mind. One of them I think is very far-fetched and is kind of playing into the future. The first one I think is more in the real world, which is agent-engine optimization. The same thing is SEO, which is how do you just make your products more discoverable by agents?

Gabe (17:12) Yeah.

Tony @ Moonpay (17:22) And I think that's very real, right? It's like, hey, the first person to crack, the first products rather, to crack SEO for their product and make it very easily discoverable by agents are probably gonna win the first wave of transactions. And I think that's very easy. I think the second piece is like, this is... ⁓ not imaginary, but I don't know if we're there yet. It's like, hey, these agents are gonna have the full picture of you. because I have like a budget and spending calculator, ⁓ spin up on perplexity computer, it has access to all of my checking and my checking and credit cards. ⁓ It'll understand how much I spend money, where I spend my money. And I connected my Amazon to it. So now it knows what I bought in the past. It's able to say like, hey, Tony, I know that you love your Bustelo ⁓ espressos in the morning. I mean, you drink about three to five of those throughout the day. Here are like three to five very specific coffee cups or coffee mugs or espresso cups that you're going to enjoy. Let me find you one that is like ergonomically designed and has cool patterns because everything you buy is like super funky. That I feel like would be the perfect world. Are we there? I don't know. How far we are, I really don't know. But my thesis is like, this is going to happen very quickly. So it doesn't matter as long as we build the blocks to do it. And the blocks to do it are like connectors, smarter models, and user willingness.

Gabe (19:09) Yeah, yeah, yeah, I think that's a good characterization. I have no idea either. I think to speculate, there probably has to be some type of distribution rail. Maybe there's a single service that does some curation or something like that, and then you can be discovered through the curation, like a marketplace in itself. So yeah, that's the...

Tony @ Moonpay (19:15) Yeah, yeah, yeah, yeah. Yeah, for sure. you could imagine like human curation for agent discoverability could be like very real. But once again, like anyone who says they have the answer, I think is being very optimistic. So we don't know. It's OK to not know. But once again, you have to build the building blocks.

Gabe (19:54) Yeah, and it seems like it's kind of something that you guys are doing, With the CLI and or with the SDK for the MoonPay CLI, you guys have some other tools in there and kind of like tricks and skills to try to hit the full package. Is that kind of in the same direction of this is like, okay, one stop shop and you can kind of hit everything.

Tony @ Moonpay (20:11) Yeah. Yeah, exactly. It's really under that principle of like human curation for agent discoverability. What famously we try to say internally is like, we want to be the opposite of MoteBook where it's like this gigantic aggregator of a ton of skills and 80 % of them are malware. What we want to say is like, hey, we have 50, 100 skills. Each one of them has been vetted by us and like we say these are okay. So we're curating this marketplace. We're curating this marketplace with all of these different skills that are verified. You as a user, you as an agent can say, hey, I want to build this like Neo bank that has virtual accounts, deep research, virtual cards, yield, like perks and prediction markets. The agent is able to say like, hey, found the MoonPay agents marketplace. Here are like the 10 to 15 different skills that you need building all of this. That at the end of the day is I think, going back to our earlier part of conversation, was a difficult thing for us at Grafane, which was like, hey, I have to build all these different agents and figure out how to get them to orchestrate together versus now within the CLI. As long as you as a human curate the right skills, agents are able to discover them and then execute and build whatever the intent is for a user.

Gabe (21:42) And how do you think about routing between the request and, because as I think about this, it's like the, you have the kitchen, which is like all these MCP might be like the stove and then you have another MCP, which is like the oven and you have this built out kitchen with all the connectors, right? But.

Tony @ Moonpay (22:01) Yeah.

Gabe (22:02) you know, to move across these different tools, maybe there's more optimal routes, right? Almost like, you know, order flow type situation. How are you looking at that? And ⁓ yeah, I'm curious your thoughts on that piece.

Tony @ Moonpay (22:09) Yeah. Yeah, for sure. ⁓ So I don't want this answer to come out like a cop out, but I do think it's like a really easy, it's a really easy mental model for a lot of these LLM products is the LLMs are getting really smart consistently. And how we think about routing, it isn't necessarily adding these different pipes, ⁓ which increase complexity and take time and so on and so forth. It's actually relying on these foundational LLM models. and allowing the LLMs to figure this out themselves. ⁓ And the tricky part there is like sometimes an L and this has been a constant theme throughout. It's like sometimes LLM will confuse swap with sale. So instead of swapping from token to token, it'll sell the token completely. Those are small errors though. And I think like the overall answer is like we allow the LLM to do this, right? And there is one large There is a Michelin level chef, which is the LLM. And there's a ton of sous chefs, which are the like tools that help the LLM, help this Michelin level chef figure it out. And I think the goal there is like, or the objective is like, hey, let's have this like the best possible LLM possible, which is the Michelin star chef. And then like in the backend, let's make sure. All the tools are set up correctly. Let's make sure the skill MD is like very, very clear. And what you have to do is very, very clear. So the sous chefs, these like tools and skills can do their job. So that's how I think about it personally. And I think that philosophy as well is kind of carried over from what we were building before. now.

Gabe (24:09) And yeah, like there as as this stuff is coming out, it's so hard to keep up with regard to like the innovation and the direction. And it's almost like a lot of people can or a few people rather in the space kind of see where things are unfolding. And I think that's where the value is, where, you know, lot of the builders are building for this centralized orchestrator that and and they become the connectors, right? They become the the transaction interface that that has other recommendations.

Tony @ Moonpay (24:27) Yeah.

Gabe (24:38) similar to like what you're building or you know, it might be like a news data, right? Maybe something that that Shoal can provide but in the bigger picture, that's kind of difficult to put together in the first place, right? But once we kind of have that foundation, what do you think is next? And with this kind of orchestrator thesis and connectors thesis, what is kind of the next phase that rolls out? I mean, we've seen like the main which agents with Anthropic, agents running their own agents? ⁓ Yeah, curious, your thoughts there.

Tony @ Moonpay (25:11) Yeah. Yeah, well, to keep it simple, think we get back to this like chat, own chat interface. That's what I think the ultimate state of it is. I don't think it's gonna be very generalized where I think it could be generalized, it could be verticalized. I think if you look at the history of software, you've gone from like extremely horizontal in like the 2000s to extremely vertical in the 2010s. And maybe AI replicates that where it's like, hey, we start out with like very generalized models. These ⁓ generalized models create like general applications. And then I think as just like software and mobile did, it finds its footing and finds the infrastructure. Then you're able to create like verticalized experiences. ⁓ And so I think the ultimate version of this is, are like verticalized chat first experiences. ⁓ So much similar like how I said earlier that perplexity is especially financed and ⁓ perplexity is extremely focused on finance. I think that's what we'll see with these like chat experiences, which is there's going to be a agent experience that's solely focused on crypto trading. What that then means is like crypto trading is a large bucket, includes perps, includes meme coins, includes Prediction markets includes yield, includes X, Y, and Z. ⁓ So I think that's what the final stage of it actually looks like. ⁓ How we get there, think like number one is like one click deploy agents. ⁓ Open call is a pain in the ass to set up. And that's just the reality of it. And I don't think people are gonna do it. Poke.com has done an awesome job of like phone number, agent, and iMessage.

Gabe (26:48) Yeah. You

Tony @ Moonpay (27:11) interact with agent. Number two is strategies or recipes, which is like, Hey, here, like, here are these like human curated strategies and skills. Here's how they've been turned into strategies. And now you can power your one click deploy agent that way. Then number three, think is like real world usage where sure. It's going to be super valuable that my agent can transact on Amazon. It's super valuable that I can buy these baskets of investments, but what's really more valuable is like it says, Hey, You love pizza or hey, you love pizza. Like I'm gonna order you a coffee beans like every single, every single week or whenever you get your paycheck. It's more valuable than my agents are usable in the real world. And so cut it up. Those I think are like the three steps to get there. And where we go is actually these like verticalized chat first experiences.

Gabe (28:05) Got it. Okay, so you think you're seeing this convergence in the longer term back to specialized ⁓ operational orchestration workflows.

Tony @ Moonpay (28:12) Yeah. Yeah, exactly. And like you can still see it in SaaS today, right? Like SaaS is inherently ⁓ a category where Salesforce itself does a lot of different things, but the sales team is cut up into different verticals that they can specialize in. And each one of those has a different sales playbook because it's inherently solving a different problem for a different customer persona. but they package it all as like one piece of ⁓ laws you to cross up, so et cetera, we've software has gone from horizontal to vertical. Now, if we say LLMs are still like very similar software, great back to horizontal, right? These generalized language models just kind of do everything. And I think we're still seeing some, we're to see the convergence back into vertical. And I think we're already starting to see that right where it's like, ⁓ A buddy of mine, Jackson, is working on surf.ai and surf is building like financial, ⁓ crypto first LLM models and a crypto first UI. So it's just a verticalized LLM built for crypto. And that's where I think this cycle kind of continues to go. Everything is sick.

Gabe (29:37) On the enterprise side, is kind of hard to get signal because a lot of these companies, especially like Anthropic, right? All this stuff is, even though they're selling to the businesses like Codex, Cloud Code, on both OpenAI and Anthropic, they're like individual contributor tools.

Tony @ Moonpay (29:44) Yes, sir. Yeah. Yeah.

Gabe (29:56) Cowork is kind of like that evolution of that, right? Which we eventually see, you know, teams can work, cowork teams, right? But on the enterprise side, what are the requests, not to get the specific companies, right? But what's kind of like the trend if you can share of what people are asking for and like, what is the nature and the shape of what they're asking for?

Tony @ Moonpay (30:19) Yeah. So I think ⁓ it hasn't really changed, right? It's like, hey, I want to create an agentic first experience for my product. I think that's like the flat out ask, right? How it happens, I think is it can vary. ⁓ But that's the overall, ⁓ that's the overall ask. Then it actually then becomes a question of like, what value does an agentic first product bring? for those enterprises and figuring out as a business how you match your product experience or your product offering to that demand. So I think number one is like expansion of TAM. That's where stuff like X4.02 becomes really interesting where it's like, hey, maybe my core business is built on $250 a month, $2,000 a month subscription. And so I've been missing. 35, 40 % of the market, which has allowed competitors to like go down the stack. Now with X4.02, I can widen my TAM and bring in those users. I think that's number one. ⁓ Number two is like diversification of business model where Claude and Perplexity and Chaggibiti do a good job of this. It's a nice blend where they have the classic SaaS model. It's like I can pay $20 a month to get access. But then I actually have this pay as you go model. And those two, don't think, go away. But then it becomes a question of what's the value metric? ⁓ The value metric I think ultimately becomes, which going back to diversification of business model is like. You charge based on completion of jobs, which is, hey, I'm doing this like data API or data product. Every time someone successively executes a trade using my data, that's the only time that they charge or that we charge for using that API. So that's, think, the second piece of like diversification of the business model. ⁓ And then I think three is like, does it enhance the consumer experience? So. Back to like headless merchants, think agents are the new websites, agents are the new interfaces, ⁓ and figuring out how to create an agent that is proactive and guides the user through your product is ultimately what businesses want, right? So I think what these enterprise requests is like, hey, I just want to create an agent ready product experience if you work backwards from it. What really means is like, are the goals? Is it like expansion of TAM? Is it diversification of business model? Or does it like amplify your product experience?

Gabe (33:18) Yeah, wow. The diversification of business model, are you already seeing teams basically charge based on results? right now we kind of went from the subscription to the token to now, I mean, I feel like we just are. Cleaning up around in the token like we're just getting kind of familiar with token base usage or metered usage, right? Especially with extra to an MPP, but now it's like, okay, you know did the service provide value is essentially kind of this next Yeah, yeah, are you seeing that people are already deploying these types of approaches?

Tony @ Moonpay (33:42) Yeah, yeah, yeah. Yeah, yeah, yeah. Not specifically in crypto. I know that this is like very much a thought experiment in the AI SaaS world, but I don't doubt that it's going to come out very soon. And it's going to be made, it should be adopted for specific products for companies.

Gabe (34:09) Okay. Yeah, yeah, because that's almost like, know, prove you got to prove your value.

Tony @ Moonpay (34:21) Yeah, exactly. then, you know, like, it comes into like a gigantic, it creates a gigantic question and a gigantic vacuum of like, if I can't inherently charge a blanket fee or blanket price, I'm going to be missing out on revenue. But if you're charging a blanket fee or blanket price, you're more susceptible to churning customers. which means you have to spend more on trying to keep those customers and you lose a lot of money when you lose a customer and you spend a lot of money trying to keep that customer. So the inverse is I actually charge for when I prove my value, which then means I'll be able to retain customers a lot more, reduce my spend and be able to iterate a lot quicker with customers who find value from what I'm doing. And it makes selling a lot easier, right? Which is like, Hey, here's like the amount of jobs that we've got done. And these are the gigantic value props for like these software companies, which is like, well, we have like a 99 % uptime. We have a 99 % success rate. like, just charge on that then, right? Cause now you're proving your value, that your infrastructure does the job for the person you're selling it to.

Gabe (35:33) Yeah. And for you guys back to the enterprise, please share what you can, ⁓ the customer segmentation, are you seeing a lot of non-Web3 customers that are looking for these types of payments to kind of future proof?

Tony @ Moonpay (35:59) ⁓ Yes, I like I so there's a couple parts to MoonPay, MoonPay itself and MoonPay has a ⁓ great ⁓ ventures team that is focused ⁓ a big strategy on acquisitions ⁓ and Iron which is a ⁓ virtual accounts and stable coin issuance company was acquired by MoonPay about a year and a half ago and We've gone live with Deal, which is one of the largest payroll providers in the world. And what we're seeing is like, or what we're hearing is, hey, we understand stable coins are a net new way to give individuals access to the US dollar. And individuals who live in places outside of the US really want to have their earnings come through that, through currency, through USDC, USDT. And what we see is like as enterprises become keen on stable coins, they choose to go to iron to figure out the strategy. And when they come to iron, it allows the opportunity to expand MoonPays offerings into those enterprises.

Gabe (37:22) interesting okay that makes that makes a lot of sense so really there's like a it's almost like an education like once you kind of do the acquisition then there's like education the natural flow is to these on chain projects to products yeah that's cool man

Tony @ Moonpay (37:37) I think so. I like if you look at if you look at ramp right like ramp is ramp is stepping into stable coins right and I think why did they start in stable coins maybe it's because what their product actually is but there's a ton of different things they could have done right they could have they could have done a ton of different things they could have done yield ⁓ they could have done virtual cards which they did but they decided to start with stable coins and so I think stable coins are an interesting value prop because companies want to hold them on their balance sheet. But also like they have gotten a lot of regulation and there's a lot of clarity. Thanks to the clarity act around what you can do with them. So it's a very welcoming asset for ⁓ companies to explore.

Gabe (38:28) Yeah, Yeah, just as we're wrapping up here, just kind of a couple last thoughts. When it comes to like tokenization, right? There's like a huge, I think there's parallel kind of narratives right now that are huge and that have revenues backing them. You know, X4.02 and these, know, Stripe coming in with NPP is a lot of traction and potential customers coming in. The parallel tracks are like perps and tokenization, right? Tokenized stock.

Tony @ Moonpay (38:37) Yeah. for sure. For sure. Yeah.

Gabe (38:56) For the tokenized stocks in RWA cohorts, and stable coins too, right? There's huge regulatory ⁓ arbitrage and bottlenecks. Like some companies will do 10 times better because they have the licensing and they focus and spend the time on the licensing.

Tony @ Moonpay (39:13) For sure.

Gabe (39:15) On the stuff that you're doing specifically, and there's probably some connections there on the stablecoin side, but are you seeing a lot of resistance from enterprise companies due to licensing and comfortability? Or is it more green grass and you're moving?

Tony @ Moonpay (39:36) ⁓ So MoonPay is one of the most licensed companies in crypto. ⁓ We have our New York Bit license, legitimately only a small amount of people like think Coinbase and Robinhood may have, ⁓ where we're regulated to the core where ⁓ I think actually to turn your question around, ⁓ enterprises and users choose us because we are so heavily licensed and heavily regulated and we stay compliant. ⁓ And so I think ultimately when you look at stuff like an Exodus, ⁓ which we launched our stable coin, the reason they chose us is because we are extremely licensed. And then that's one of the big value props for MoonPay as we were thinking about joining is The big ethos here is how do you become backwards compatible with the existing financial system? Because what MoonPay wants to do is build the operating system for modern money movement. And I think you can do that in a very cowboy style. But the way that you scale out and get big wins ⁓ is by being regulated, having your licenses, and being extremely compliant.

Gabe (41:03) Cool man, yeah, this has been super helpful. All things agents and CLI and MCP, it's a lot to wrap your mind around. I think it's difficult too, especially online, to kind of see what's real, where are the trends actually settling, what are the institutions doing. So yeah, thanks for coming on and appreciate you joining.

Tony @ Moonpay (41:25) Yeah, of course. Appreciate you for having me, Gabe.

Gabe (41:27) Yeah, likewise. All right. Take care.