How AgentCard Works: The Visa Card for Agents by Alchemy

For having me on.

We're we're really excited to have you. Excited to hear a little bit about your journey and unpack what you're building now. I guess with that, I wanna kick off and jump back to even pre alchemy, pre agent card. You were building Heymet for a few years. Tell us a little bit about that and then the acquisition and how you got to sort of the place you are now, and then we'll jump a little bit more into agent card.

Totally. Yeah. I mean, I've been a big fan of NFTs NFTs for a very, very long time. My company before that also doing NFTs. At Haymit, we really firmly believe that the last thing we needed to emulate the physical art market at a digital level were NFTs. Right? They're that final unlock that allowed to properly own something on the web. So we end up working with a lot of creators to help them launch their own collections and basically start making money online. Right?

A lot of these people beforehand would have to place their work in art galleries, eat through terrible fees, or they would have to go to art fairs. And this, like, NFT thing was a huge unlock for them to go out and leverage their online distribution and start selling these things that people truly owned online. So we end up working with a bunch of really cool artists, a bunch of really cool brands to sell, I think, something around $38,000,000 of NFTs.

And our mission's, like, very much aligned with Alchemy from the start. We had known Nikhil and Joe for a while. We were trying to bring NFTs to as many artists as possible. Nikhil and Joe were trying to bring web three to as many people as possible. So it seemed like a very natural fit, we realized that we'd be strong together and that it'd just be a huge force multiplier if we joined forces.

And that's how we landed here. Now we're at Alchemy. It's awesome. Yeah. I love it. What I love it. What were some of the core pieces of tech that you think got them excited or, like, the muscle that you built at the company that ended up being the thing that really they were buying in this? Like, I'm imagining some of that core underlying infrastructure that you guys were getting volume through and using was was pretty interesting to them. What what what were what do you think those kind of areas of payment?

What what was that?

I think there are lot of, like, very interesting pieces. Right? Obviously, Alchemy, quite active in NFT space. We have a lot of products like your NFT API where we serve all kinds of data related to NFTs. We used to have some other platforms in the space very similar to what we built at HeyMint there, which were definitely interesting to Alpen. I think Alpen had a really great idea of the technical challenges that came with building out some of these systems.

But I think, honestly, the biggest thing that is of Alpen was just our ability to execute zero to one. We've gone ahead and built all these different platforms as part of the Hayman umbrella. And now at Alchemy, our team is doing the exact same thing. That that's really the magic sauce.

I love it. I love it. I'm I'm curious to hear more about that because I am flashing back to a conversation I was having with Drew maybe in February. We're at a very stable conference in San Francisco, and people you know, it's like, what are the buzzwords in the room? One of them was about Agentic, AgenTic Finance, AgenTic Commerce. What's going on here? And I was kinda feeling now, like, okay, what's real here? What's really happening? You're talking about bringing things zero to one.

It seems like with Agent Card, you've done that, or you're in the process of doing that, at least. Tell us a little bit about, like, what it's like to make agentic payments real, and, like, what's really happening here. I love the website you guys have, like, buy DoorDash, book an Airbnb, like, all this stuff. Tell us about where we are now with agent card and, like, I guess, how real it is. Like, how far are we on the zero to one to get towards actual adoption of agentic payments here?

Totally. I mean, we're seeing real adoption. Kind of zooming out a little bit, what we saw is that AI is really good at deciding what to do. Right? It can come up with these really elaborate plans. It can browse the Internet. It can do all of these things, but it was missing a few pieces to actually go out and exist and act and transact online. And that's what we do at Agent Card. We enable your agent to actually go out and act on those plans that it creates.

So we give them access to payment methods where it can actually go out and buy all these things on the Internet that already exist. Previously, there are a bunch of issues that you could either shop at a very limited set of companies that supported agentic payments, which there were very few of, still are, or you could go out and give your agent access to your entire credit limit and hope nothing went wrong. We at Agent Card kind of saw those problems, and we're like, we need a solution here.

We actually gave our credit card to one of our agents, and bad things happened. So we decided that we really needed to build a way for agents to exist and transact securely online. That's what the idea from agent card came in.

Love it. I'm not gonna ask what the bad things were, but I am curious, but we won't ask that. What is what has been you know, maybe other than that, what something that has surprised you in terms of getting this to market? What has the team learned? What what seems to be, like, an interesting part of this build as you've both internally built it and then also seen some of the companies or early users of this thing? What surprised you most?

I mean, I think the use cases just surprised us tremendously. Right when we launched this, we were met with this huge crowd of people who are nontechnical, don't have a strong technical background, who have gone ahead and basically used these harnesses like OpenClient or Mez to go out and automate entire businesses and start making a bunch of money by just automating all these things.

Actually, a lot of our initial traction that we got was with people who were just looking to automate a lot of their business needs, whether it was procuring products, doing a bunch of other things. That's kind of where we saw our initial traction coming from.

Yeah. That's interesting. I was gonna press on that. Like, pain is this really solving? Like, what's the why behind it? Is it that kind of recurring transactions right now, or, like, what do you see as the immediate pains that are getting solved here?

I think there are a lot of very interesting use cases that we're seeing pop up. I think we're shifting from a model where you talk to the traditional ChatGPT interface. You define exactly the outcome, and you have it give you some response, right, to a model where you're defining goals and agents are going on acting pretty much autonomously towards that goal. Right? Previously, you might have asked it to go and look up what the current World Cup score is.

Now you're asking it to, like, go and find, I I don't know, the best hairdryer, for example, and just buy it for me. So what we're shifting towards is a world where you're really just delegating outcomes. And for agents to be able to act on those plans that they construct, they need a bunch of different elementary things. Biggest ones are identity and payments. Right? They need a way of existing online, I.

An email address, a phone number, and then either way to actually, like, go and take action, go and pay for things, which is what we're building here at Agent Card. And we've integrated all those other parts of this stack as well.

What were some of the things that you found connected? It sounds like the use cases were surprising, but what were some of the use the the connective themes that you saw of the ones that are working? I think some of question behind the question here is some people are excited about this technology, but in the show, something we're really pumped to explore more is, like, what is actually driving volume and why might that be working?

So what for those things that are ticking off that are a little surprising, are there any themes that you're seeing of, like, types of problems they're solving or specific pain that is going away or any learnings in that sense thematically?

Yeah. I'd say I'd put them into two different boxes. There's obviously the prosumers who are really excited about this new technology and will adopt a new technology to, like, go out and automate some arbitrary Right? So we have a bunch of people who are basically trying to automate their entire life, trying to automate buying groceries, all these different things, which are really cool.

But then we're also seeing another segment of people who are automating real problems for their businesses, which I'm very much interested in. We're seeing people automate these recurring procurement processes and all of those kinds of things. And so the real needs were met. What is a procurement process that's getting automated? Can you dive a little bit into that? Totally. Yeah. It really depends on the company. Right? Here at Alchemy, we have this thing where every week, we try to stock up our office with snacks.

Right? That's a perfect example of some of those recurring procurement processes where you have to access some platform that's not built for agents. It's built for humans, and you don't want to spend the hours it takes every single week to go out and do that task.

Yeah. I'm curious about that thread of, like, automating life, automating the operations of a company here. Like, some of what I'm hearing is maybe back in the day, you would have outsourced this to the office manager or the administrative assistant or even, like, an overseas virtual assistant. Right? Like, we were it's not novel to say, like, oh, we're gonna delegate our sort of manual tasks to someone else. Obviously, this is a very novel acceleration of that.

Where is it like this is just replacement of human tasks with a machine? And where is this like, no, this is a whole new set of capabilities. Like, we're not just talking about human intelligence here. Like, agents are gonna do it differently

than your assistant would have done it. I think a lot of it is saving time. Right? I don't have the or we at Alchemy don't have the time to go through the entire catalog of DoorDash and take the best snacks that are coming out. Really, what we're doing is saving a bunch of time for a lot of people and making these things a lot easier. I think that that's the biggest thing we're seeing right now and the biggest drive of value. And and just like agents don't sleep.

Oh, I guess. Agents don't sleep. Agents never sleep. Sleeping takes a lot of time. That's for sure. About half our lives. So that is that is a big one. Speak a little bit about the Visa relationship and where the Visa card comes in, how that how that's bringing to life some of the utility side of this.

Totally. Yeah. What we realized very early on is that it's incredibly hard to give agents safe access to the traditional economy. Right? These what these major retailers like Amazon are not going to adopt stablecoin reels overnight. It's going to take a very long time before we see kind of a critical mass of agent reels being integrated. So what we really need to give to these agents is a way to transact in the real world, And our partnership with Visa allows us to do exactly that.

We're able to take in an existing card, whether it is from Capital One or another bank you might be with personally, and we can create secure virtual cards from that that never expose the actual card details to the agents and that have some limits on there, whether it is spent limits, arbitrary amounts, whether it is category lock, merchant locks, all those different things we're able to enforce on these derived virtual cards from your normal card.

Yeah. I would love to this is really interesting. We'd love to dive into some of these weeds a little bit so you have I think this is where where it gets really interesting. Like, who's managing these limits? Who gets to set them? Are those default? We something we're also exploring is this idea of, like, the merchant of record of these transactions. What can you share surrounding the role that Alchemy is playing versus, like, the tooling it's giving? How do what does that look like?

So enforcement of all these limits is happening at the actual payment processor level. Right? We're not enforcing any limits. We allow you to set them, but who's actually enforcing this is Visa. They're the PSP here. They're going to or they're the network here. They're going to actually go out and enforce those limits. And the motion of record works exactly the way it would typically work. Right? We're not man to middling these transactions through a partnership with Visa.

We're able to basically leverage the infrastructure they've spent many, many years building up and use that exactly as it comes for agent applications. So in in the, you know, unfortunate in the unfortunate,

like, you know, let's just say reality where one of these people who's using an agent or they're acting in some sort of, like, illegal way, what would that who who would take the l for that specifically? You're saying that's a visa problem or maybe the like, who who's on the hook there in in in that front?

I think it very much depends on the type of abuse we encounter. What I'll say is I'm obviously not a lawyer. It very much plays out in the way that you would see it traditionally being played out. If someone is using an agent to do various things on the Internet, odds are they're not going to have a great time. It's not like there's another another component of liability being introduced there. It's very much distributed as it was before.

What have you specifically learned around this part of the conversation as a builder and founder? Like, how are you thinking about these maybe common landmines or things you wanna avoid? I feel like a lot of people are excited to build in this space, but there's also, like, I think a healthy worry around stepping on landmines and doing like, you like, having agents, you know, be kind of off off into the frontier doing things.

How are you thinking about that in terms of the risks and rewards of of how you're building into the future?

That's a really good question. I've been very fortunate to have been thrown into the deep end of financial services as a whole and kind of have been forced to explore the entire landscape both through what I previously worked on here at Alchemy and what I'm currently working on in the a d department. It's very interesting. Right? I think you really do need to do some, like, elementary research on how these financial services traditionally work to understand where the risk lies.

There's really no way of getting around it. If you're trying to interact with traditional financial systems, you're going to have to do a little bit of homework to learn how they work, where the risk is managed, and how that all works.

Earlier, Flor, Drew was talking about the sort of volume of adoption here. So is what I'm hearing you say is, like, if I wanna go on the user journey with have AgenTic payment integrated in my life, and this is any person or company, I can go create an agent using agent card right now, and my agent gets an email, phone number, and can start making transactions.

When you think out, like, six months, twelve months, like, maybe even years, how do you imagine that sort of is it, like, everyone is going to have this sort of agentic assistant making purchases? Every company is going to run their supply chain through this. Like, what is the adoption if if you get this right and other players in the space get this right? Like, what does major adoption of this look like?

It's a really great question. I think if it will follow a very similar adoption curve to the adoption curve we're seeing agents follow. Right? Like, obviously, basically, every single tech company today is exploring either exploring, adopting, or already adopting and productionizing agents in some capacity. I think having them actually go out and act and transact online is a very natural extension of that.

So if the future plays out the way you want it to, every company will probably be transacting in the near future. I think we're shifting to a model where delegation is the key. Right? We're not trying to use agents to create plans for humans to execute execute anymore on anymore. We're trying to have agents actually go out to create those plans and then complete them and create a full circle moment where they can deliver the outcome that you're looking for.

So to recap your question, yes, I think this can be very big. And, ideally, we have every single company transacting very soon. Hey. I mean, that's where it gets really interesting when it's, like, you have sort of the agent strategizing as well. It's not like, hey, please buy me Nike Zooms in size twelve from this store. It's like, no. Like, give me running shoes for the $120

budget. Like, all this stuff. Yeah. The reasoning. Maybe, you know, I I'm imagining, like, someone like, an agent started to know your life the way that these LLMs know our lives if you're using them a lot anyway. Like, sometimes it's creepy. Right? But, like, Claude or Tatchy PT, like, knows my wife's name, like, knows things about me, like, personal and professional. And I'm kinda cool with it because I want it to be personalized. Right? Like, that could go on a level. Like, it's like, Zach, like,

how do people get involved? What are some of these ways that people are typically getting started with this? And, you know, if if I'm a business right now, I think a lot of companies a lot of folks who are listening to this show are either founders or operators in enterprise. So they're building their own companies. They're working as a professional in this space. We're definitely not probably reaching the kind of retail, like, consumer use case.

So how would you position what's possible in a very basic way to, you know, a a a professional who might be building a startup or an enterprise company in this world. What are some easy ways or they they could get started using this system? Like, a low, like, a low risk way to get going.

Totally. Yeah. I mean, if anyone is currently purchasing things manually on a recurring schedule, I think that's a really great way of just getting started. Right? If you already have an agent spun up, what you do is you just give it the instructions we have on our landing page. They will sign up for an agent card itself. It will basically ask you a few questions, ask you to put in your traditional credit card securely, put in a few other details, and it will be able to transact online very simply.

If you do not have such an agent running yet, there's a lot of real cool platforms that allow you to spin up an OpenClaw or her mess agent within seconds. I think Cloudy is one of them that comes to mind. I think that'd be the best way to kind of start experiment experimenting with it.

Also, if you live in San Francisco, a great use case actually is paying your parking tickets. I've been doing this a lot. There are many. Pictures in San Francisco. Agents, and they will figure it out. It's great. That's amazing. So you literally just take a photo, agent looks at the details, goes to that site or whatever. Because it's always, like, you know, a manual site you're putting into your phone.

Yeah. It just figures it out. Has a browser, has an agent card, just goes and paste for it. Correctly in the Claude or Chachi Pitita or whatever model you're using, like, you're using that app, and the agent card is authorized through it? Yeah. I talked to my OpenClaw on Telegram.

So I just sent a picture and figures it out. Are you finding that Telegram is the best, like, most accessible or inclusive messaging channel to I think that's another interesting question is, like, where do we actually engage here? There's, like, the agent itself, but then there's the chat or the application layer. Or so it sounds like you're using Telegram, an agent under that, and then that agent's provision to go to these sites on your behalf.

Is does Telegram and the messaging like, what other options exist outside of Telegram that you're seeing for places you can go to engage with your agent?

Yeah. I think people have varying preferences based on how they're using their agents. We're seeing a lot of businesses wanting to access their agents through Slack. As an example, I think most individuals, most consumers are going to want to go with something that is, like, fairly easy to set up to set up, which is why we're seeing Telegram get a lot of adoption. Obviously, it's very easy to set up a bot.

On Telegram, it's little bit harder to go to iMessage route and set it up with a phone number, do all of that where it can do iMessage to its full full extent. I'm kind of in between. I like about Telegram. I like iMessage. IMessage sometimes does not have a good time, so then I go back to Telegram, and I stick to it.

IMessage would be my preferred route, I think, if I mean, I'm I'm an iMessage Maxi, I think, unfortunately.

Yeah. I think you really have to meet people where they're at. Right? You have to meet them where they work, where they interact already. Apple needs to acquire Whisper Flow so that the the voice to text is much better. And then we'll get it integrated with iMessage, and then we'll be doing agent to commerce with an actual good, like, voice command interface. Be much better. That'd be incredible.

It would be cool. Well, this is fun. Flor, you're our first ever incline segment. So segment guests, so thanks for being on this. And, also, an open open invite to invite others. I'm sure you probably are connected to this is the first day we're getting this this part of our show off the ground, really focused on this topic.

So if you have any other builders or founders that you admire, you wanna give flowers to, would love to have them on the show, talk about what they're building, and continue to help people make sense of what what is now possible with some of this stuff. So appreciate you being a little bit of a guinea pig today. Totally. Thank you so much for having me.

I learned about twenty minutes ago. Love it. Love it. Alright, man. We'll have you back soon. See you. Thanks so much. Take care.