Agentic Trading Unlocked: Nansen AI’s Onchain Intelligence Layer

GM, GM! Welcome to what is going to be another fantastic episode of DevNTell. If you didn't know, DevNTell is a 30-minute podcast held every week allowing founders, hackers, and anyone in between the opportunity to come on the podcast and showcase what they've built. Today, I'm ecstatic to welcome Hürcan Polat, who is the growth lead at Nansen. If you don't know, Nansen is a blockchain AI analytics platform that enriches onchain data with millions of wallet labels. If you stick around for today's episode, you'll get to meet Hürcan, learn about Nansen and how you can get started using it today, as well as an engaging conversation around AI agents and agentic trading. All right, let's get into it.

[DevNTell Intro Music]

GM, GM! Welcome to the show Hürcan. I'm ecstatic to have you on today, man.

Hey Narb, good to be here, man. How are you?

Doing well, doing well. It's Friday, another episode of DevNTell. I'm excited for the topic today, and I'm sure our audience is as well. Agentic trading, AI in general, agents—super hot topic, especially this year. Really excited for our audience to learn a little bit more about yourself and Nansen. So, I guess with that, would you like to just give an introduction of yourself?

Yes, sure thing. Hey everyone, I'm Hürcan. I'm leading the growth of our developer products here at Nansen. I've been with Nansen for close to two years now, and since last October, I'm responsible for anything related to Nansen API, MCP, and CLI. This includes partnerships, integrations, developer relations, growth initiatives. So basically, if a dApp developer, a fund, or an AI agent wants to tap into our data systematically, that would be my world. I also act as a bridge between the product, go-to-market teams, and our customers. So, if there's any feature requests or any bug reports from one of our users, I act as the voice of the customer as well to get stuff done.

Amazing, amazing. And just curious, how did you find yourself in the tech scene and eventually find your way to Nansen?

Yeah, so honestly, when I was at university, I had zero interest in crypto. Obviously, I was aware that it existed, and some of my friends were into it, but it wasn't really on my radar. I studied metallurgical and materials engineering, so it's very far from either software or finance. So, after graduating in 2020, I worked as an engineer for a while, but then I decided it wasn't really for me. Around that time, I guess the Turkish Lira was declining faster than usual. So, I knew that I had to do something else, and I slowly got interested in crypto. Nothing crazy initially—it was just like following a couple of KOLs on Twitter, learning how to do some technical analysis, stuff like that. But I think what got me into the sector was I was active in a couple of different Telegram groups, mostly like trenches, trying to find the next mooner. Through there, I met some people, I did some networking, and one of those guys said, "Hey, do you want to do something part-time with us?" That's when I pivoted to crypto, left my engineering job, and since 2021, I've been full-time crypto. I worked at a couple of different protocols before Nansen. I started in a very small layer-1 project called Lamden, then I worked on another data analytics product. Afterwards, I worked at a Perp DEX, and then finally here I am at Nansen. As I said recently, I've been focused on the API and agent side of things.

That's quite a journey, man. I kind of resonate with that as well—I got into web3/blockchain around 2021 as well. So, we both experienced the same kind of hype cycles and bear market. I don't know about you, but what we're kind of going through with AI now sort of feels the same as how crypto felt. There are so many projects out there, so many different things going on at the same time. It's hard to keep up. Since the start of the year and towards the end of last year, we're seeing a lot more building blocks being put into place to help power agentic trading. The models are much better, there's better tooling now. From your perspective, where are we at with that overall scene, and where do you think it's going to go in about a year's time?

I think we are still laying the rails right now. I mean, especially since OpenAI, Claude, and other well-known AI agents came to presence, we are seeing much faster progress because now people understand how different it is to use an agent versus using an LLM, and the capabilities they have. As you said, in the last year we've seen a lot of big companies and institutions backing both agentic payments and commerce, like Coinbase pushing for AI wallets, Stripe pushing virtual payment protocols, and such. Although there are good progress in agent capabilities like accessing the data, doing research, and paying for its own usage, I think we are still waiting for the adoption to catch up. I mean, you, me, other people—I guess we can see the destination where, in the end, it's going to make more sense to do everything with your agent in addition to trading. But at the moment, I don't think a lot of us are still comfortable handing real capital to an autonomous agent because we hear about all of these incidents with agents deleting databases, or hallucinating and such. So, I think as this technology evolves with less hallucination and better understanding of intent, we are also going to be seeing a much more explosive growth in terms of adoption. I mean, exact timelines are really hard to say because AI is kind of moving similar to how crypto moved a couple of years ago. It was really so fast; something new was happening every day, a new dApp coming into presence every day. It's similar to AI at the moment. So, I think we still need a bit of time for the infrastructure to be a bit better so that we can hand all of our stuff to those agents. But I think it makes sense to work for that future at this moment.

Yeah, well said. There are a lot of different things going on at the same time—all these different protocols, from Stripe to Google is in it too. From my perspective, a lot of the hype and products are very developer-oriented at this moment. We're not really seeing too much on the consumer side beyond chat apps like ChatGPT, but I think it's coming. I think the money is going toward the developer side, which makes sense given how expensive it is to run some of these things.

It is, it is.

Speaking of developers, we have a largely developer audience today. For those who might not have heard of Nansen before, can you introduce us to it?

Sure thing. First of all, I'm an engineering graduate, but I'm not in software, so it took me a while to actually get into this stuff. Of course, when you first see ChatGPT, everything makes sense, but once you start messing around with MCP, CLI, and those agent harnesses, things get complicated a bit. But to talk about Nansen, Nansen is an onchain data analytics platform used by individual investors, quant funds, AI agents, to do discovery, due diligence, and defense. How we do that is through our wallet labeling. As you said, Nansen has labeled more than 500 million wallets across more than 20 chains. So, through that, you're able to see not only like "0x123..." but you see actual entities like Wintermute, or Paradigm, or a smart money wallet, etc. That is the context, which is the real core product. With that labeled data, you can track how funds are moving onchain, what smart money is buying or selling, or find out who else holds a token that you're exposed to so you can understand the risks and be alerted if anything goes wrong and you need to get out of a position. So, I would say Nansen is an onchain data analytics platform, but we also do some other stuff as well. A couple of years back, we acquired one of the staking platforms, so we are now running validators on more than 20 chains. Recently, we started to go heavier on the agentic and trading side. It all comes from the labels, but the end product changes depending on the technology and what people need.

Amazing. Having all this data, it makes sense to also have it play nice with AI and agents and have them be able to consume that data. From your standpoint, when did Nansen see this natural merging of the two?

I mean, to be honest, Nansen was founded in 2020, but our initial brand was nansen.ai. So, either Alex Svanevik, our CEO and founder, had a visionary view, or it just sounded too cool to pass on. But we have been involved with AI since the early days. I would say the major changes we did started last year, as soon as we started to see a lot of activity coming in from agents. Obviously, in the world where agents need to action on themselves, they need quality data. We realized that in addition to all of these people that we have already been providing data to—like funds, exchanges, institutions—AI agents is actually a new customer category that is quite untapped by a lot of dApps or products. So, it was last year when we had a clearer view on what we want to do. Obviously, when we started with the API side of things, it was still more regular users, but then we released our MCP, and a couple of months ago, we released our CLI. It's a command-line interface tool which basically has all of the functionalities that our API has, but it's more for agentic use cases. Rather than having your agent make API calls, it's much better at using the command line to enter commands and get responses. We've added a bunch of skills to it so it's also able to understand how it is able to use and what sort of data it can get out of each product. To be honest, at this moment, when you give an AI agent any breadcrumbs in terms of what it's able to do, most of the time it's able to figure stuff out on its own. So, whether you give it our documentation website or you install Nansen skills from Claude-Hub, most of the time it's going to be able to pull the exact data with all three of these resources. So, up to now, it has been up to the developer to make the selection for its own agent.

It sounds like you have a lot of things cooking there and a lot of tools available for folks to use, which is amazing. For those listening and watching, those resources are linked in the episode description below, so if that interests you, definitely check that out. Just to double-click into the data itself a little bit—how do you guys go about enriching all this data and making it far more useful than say a regular indexer might present? What is the difference between a regular indexer presenting this data and Nansen's version of it?

I actually had a good analogy for that. You can think of raw data sources, either RPCs or indexers, as an entire city's phone book, but without any names, right? So, it's just phone numbers. You can technically call them or try to understand who they are, but it would take years of work to understand who's who. Nansen's labeled data is the same phone book, but with names, job titles, reputation scores, and credit scores attached. For an AI agent, the difference is huge because instead of seeing "0x123... transferred 500 ETH to another address", the agent is now able to understand that a tier-1 VC fund rotated out of ETH into SOL positions, which gives a whole other context to the agent, which is what agents need. In addition to that, it brings cool stuff to the agents, which they desperately need. First is entity identification—who is behind this wallet? But in Nansen, we have much more labels compared to that, like behavioral classification, which I really like. For example, is this wallet a smart money, are they a whale, are they a retail investor, and what are the sectors they invest in? Also, historical context is really useful for agents—what has this wallet done before, who did they interact with, send funds to, or what is their PnL like within the last 30 to 90 days? These are all of the data points that an agent needs in order to make a decision. This doesn't have to be an immediate trading decision, but it can be. If you're already holding some assets, you can do due diligence on it and defend your portfolios. It basically gives you access to a broad list of tools.

Amazing. I bet there must have been a lot of challenges massaging all this raw data into a meaningful format. Off the top of your head, were there any particular "oh my gosh, I don't want to go through this again" moments that you went through, or something you want to share about labeling wallets?

Yeah, so the labeling is done by our attribution team. We have a team dedicated to labeling those wallets, and of course, we ourselves use AI solutions to understand and to make sense of what those webs of interconnected wallets are like. When I first joined Nansen, it was pretty amazing to me because obviously I knew a lot about crypto, but seeing actually "Hey, this belongs to a hedge fund, this belongs to a market maker"—when you know about this, it really changes your perspective when you're trying to comment on these activities. It is one thing when a market maker sends some funds or has an action, versus it's something completely different when a fund sends some of their assets to a centralized exchange, which we usually refer to as selling. It still amazes me to this day to be able to take a look at some random numbers and to understand actually "Hey, this is Vitalik, or this is Sailor", or whatever. I'm quite grateful to our great attribution team for that.

Yeah, 100%, they sound like they do legendary work. Kudos to them. On the topic of this data, since we are a developer show, what are some of the different ways and products/tools Nansen has available for either developers or their agents to consume this data?

Yes, so currently we have three different products suited to different customer profiles. The REST API is obviously the foundation—structured endpoints for either token or wallet data, which helps you understand what smart money is doing, what interesting wallets are doing, etc. And then we have the MCP server. I think we released it like two months after we released our API. Model Context Protocol has its use cases. Most of the people that use it connect the MCP server to their Claude, so while they're talking to their Claude, if there is anything that makes it necessary to take a look at the onchain data, Claude is automatically able to query Nansen's MCP server and then give you a human-readable response. The newest addition to the family has been the CLI. I think it was three months back or so. A command-line interface tool which basically has all of the functionalities that our API has, but it's more for agentic use cases. Rather than having your agent make API calls, it's much better to use the command line to enter commands and get the responses to it. We've added a bunch of skills to it so it's also able to understand how it is able to use and what sort of data it can get out of each product. To be honest, at this moment, when you give an AI agent any breadcrumbs in terms of what it's able to do, most of the time it's able to figure stuff out on its own. So, whether you give it our documentation website, or you install Nansen skills from Claude-Hub, or you come across our MCP server on one of these registries, most of the time it's going to be able to pull the exact data with all three of these resources. So, up to now, it has been up to the developer to make the selection for its own agent.

It sounds like you have a lot of things cooking there and a lot of tools available for folks to use, which is amazing. And you mentioned it earlier, but you guys also have X402 support, being able to pay per request for some of this data. If I'm not mistaken, I think I saw it on the Agentic Cash marketplace, or maybe it was something else. Do you want to speak to that?

Yeah, for sure. We have both the X402 and the MPP integrations. Whenever there is a new agentic commerce or agentic product coming out, we are usually the first onchain provider to actually do the integration. I've been following X402 since last year since it first came out. When it first came out, the volumes were crazy high because I'm assuming there was like 90% wash trading going on, and the X402 volume dropped significantly towards the end of the year. I guess either people didn't want to wash trade or they had another focus. But since the beginning of this year, we started to see the volume climb up again. For me, I think it's the best thing for an API to be available to agents because agents hate subscriptions, right? They don't go to your website, fill in their stuff, approve their email, generate their API keys through a web app or whatever. So, you have to find a way to give them an easier way of access and an easier way for paying for things. On our side, giving them easier access was through our CLI and MCP, and getting them to pay for it was through our X402 integrations on Base and Solana, and as I said recently, on MPP. Obviously, the X402 volume when you compare it to the regular users who are using REST API with API keys—obviously REST API is at a much better place at the moment because of the adoption it had up to now. But we are starting to see the space is starting to heat up. You can obviously see this with how institutions are investing into this and all of these big platforms like Google, Stripe, they are trying to be the first to do the things. There's also some sort of war of the standards, at my point of view. Some of the more institutional players are supporting the Stripe route, while X402 is a bit more open source. So, I'm also looking forward to seeing which of these methods, or if it's going to be something else, that is going to break up that agentic commerce.

Yeah, I'm with you. There are many different companies and foundations competing for the same piece of the pie. I think competition is always great for innovation. Eventually, I don't think it'll land on one particular thing, I think it'll be a use-case-to-use-case thing. But yes, I've also seen the uptick in X402 adoption. The only thing that is unfortunate at the moment, but I'm sure will resolve itself, is there isn't many things to buy at the moment. But I think it's because it's an evolving space.

Yeah, I would say the use cases are also evolving. At the moment, people are trying to sometimes make their agents do stuff that doesn't necessarily need to be done by agents, because for some of the stuff we're not there yet, it takes too much hand-holding, and you basically need to have a human in the loop on most of these cases. But at least it's good that compared to six months ago, I think the capabilities we have at the moment is much more superior. So, I'm pretty sure that all of these like friction points, they are going to be resolved in the next year or so.

Yeah, I think so as well. The space is moving so fast now, and development cycles are like 10x, 100x. So, it'll fix itself soon enough. On that same theme, can you share some alpha or anything around some of the things you might be releasing this year or next?

Sure. As I said, we recently launched trading on Hyperliquid through Nansen, and the reason is because we want to have trading on spot with Solana and Base, and with perps on Hyperliquid. We want Nansen to be the single stop for you to both get your alpha and also execute on the alpha. We don't want you to come check out Nansen, see what's there, find some alpha, go somewhere else, move your funds, and trade through there. We want this whole thing to be like one seamless experience. So, I'm quite bullish on trading on perps on Hyperliquid. That's our main focus at the moment. Towards the end of the year, we are going to have a deeper focus for agentic trading, and it's not just going to be like writing to an agent "Hey buy this, buy that, or before that, get to know a bit more about the asset." It's also going to enable you to create strategies and also backtest those strategies with historical data to see if your strategy actually works or not. Soon, you're also going to be able to share those strategies with other Nansen users as well. And the third thing—I don't know if we're going to be able to do it this year—but prediction markets. As you know, in addition to perps, prediction markets is like the second highest mindshare in terms of crypto. In the API, we already have some integration going on with Polymarket, so you are able to get like aggregated, summarized both market data and results through our API, but we want to go heavier on it to also have smart money for prediction markets. Rather than seeing all of these insiders on X where you say "Hey, this insider put X amount of money in this prediction market just before it closed", which you obviously can tell is a manipulation or if it's an insider, so I'm really bullish on seeing actually those either insiders or those named wallets on Nansen so I could actually see what they're doing as well. I don't know if you've seen it, but recently—I think it was yesterday—I saw that Truth Social is selling like faster access to Trump's tweets or something. So, it's crazy the type of data that is coming on and the type of insider plays we have. So, I'm hoping with Nansen we'll be able to look through some of these and make investment decisions better. So, I would say those are the stuff I'm waiting for in Nansen.

Amazing, and yeah, I did see that yesterday. It'll be interesting to see what comes of that and if other social platforms will also take that and run, because I mean, yeah.

It is interesting. I want to see what comes out of it. Either more people are going to have access to those insider tools, or is it going to make those market manipulations worse? Are we going to see Trump sharing something on Truth Social and then like open a reverse trade at that exact same time or something? Time will tell. Nansen and other data analytics platforms is the reason you need stuff like this so it's good to keep your hand on the pulse, and if there's an insider play that you could find and catch onchain, it is Nansen that you should come to.

Awesome, awesome, good to know. And none of this is financial advice, by the way. It's just data. I think the entirety of the podcast kind of summarizes this, but in your own words, for people who might have gotten bearish on the crypto/web3 space, how would you summarize to them to come back on ship?

I'm still bullish on crypto, and it's not because I'm working in the industry. I think blockchain and crypto is the biggest enabler of AI. When you think of it, what AI agents need to manage those financial decisions—it needs to be verifiable, trustless. All of these permissionless rails crypto has, I think it's going to be really beneficial for AI as well. I mean, we're seeing how agentic payments are doing or what it is capable of doing, and we also see stablecoins as like the one of the biggest winners of this AI agent race as they are accepted by all of these different contracts and they're not tied to a specific country or anything. So, I think all of this transparency, the permissionless nature of crypto is going to work really well with AI. So, I think you shouldn't leave crypto just because AI is the new hot thing. I'm not saying the bubble is going to pop, but I think there are definitely some huge valuations that doesn't make sense for me. So, I'd like to be in the place where both AI and crypto coincide and get people the best data possible so that they can do those decisions on themselves. But yeah, I think that's all.

Well said. Exactly, there's a natural fit here. That's kind of where a lot of the hottest innovations will be, beyond just model development. As we're closing in on time here, from your perspective, what's the best way for people to follow along with Nansen, get started, and are there perhaps any job openings? I know the job market is a big thing for developers these days.

To be honest, I don't know if we have any developer positions open, but we do have a positions page available, so for anyone that is interested, I suggest them to go and have a look there. In terms of how to stay in contact with Nansen, obviously the first thing would be to follow Nansen on X because whenever we have a big release, that is the place where we are announcing all of this stuff. We also have a blog and a YouTube channel. With the YouTube channel, we also do some workshops. So, for example, if you're interested in the API or how our smart labels work, or how Nansen's web app works, we have a lot of tutorials already available on YouTube at the moment. And otherwise, just come and sign up to Nansen. Whenever there is something we need to tell to our users, we send out an email to them explaining them the changes. And myself personally, I also send a lot of emails to our API users as well whenever there is a new endpoint coming up or if there is any changes with the endpoints that we currently have or anything that they need to know. I usually email them or just DM them on Telegram. So, I guess that would be the best way to get in contact.

Excellent, and all those resources again will be in the description of the podcast below, so definitely get in touch with Hürcan and the team if that is something you are interested in. And with that, Hürcan, thank you so much for taking the time out of your busy day to come chat with us today. It was amazing to meet you, converse on AI, and learn a little bit more around Nansen itself.

Thank you, Narb. Thank you for the opportunity. It was great to have a quick chat. I'm looking forward to having more in the future.

Of course, it'll be my pleasure. And with that, I just want to wish everybody a very happy Friday, happy weekend wherever you may be, and we will catch you back here for another great episode of DevNTell next week. Till then, have a good one.

See you everyone, bye.

Cheers.