New Ethereum talks, every Monday. The week's conference uploads by event, in your inbox.

Loading player…

High Throughput Gasless Payments for the Agentic Era | Bhushit Agarwal & Harshal Bhangale, Circle

Ethereum DenverMon, Mar 9, 2026, 12:00 AM

🚀 Get Ready for ETHDenver 2026! 🚀 We're already hard at work preparing for next year's biggest Web3 event! Keep your eyes peeled for more info on ETHDenver 2026—it’s going to be epic! 🌟

Transcript

Good afternoon everyone. I hope you're all enjoying your first day at E Denver. Really excited to be here. Um I'm Bushett. I'm Hershel and this is Bushett and uh yeah we're here to talk about agent e-commerce the challenges in the space the opportunities and how we've come up with a solution that we think can help scale microp payments.

So the internet for the last 30 years was kind of built for just one customer in mind right humans. Um and then we built interfaces and payment rails with humans in mind. Even the monetization schemes were catered towards humans. So you had your subscription models, you have your you had your ad revenue models. Uh but with the advent of autonomous agents, there's a new customer, right?

And these AI agents, they don't really see ads, nor do they subscribe to any tools. So, how do you monetize for this new customer base, right? Well, to truly unleash the power of and the potential of these AI agents, you need to empower them with the ability to access whatever data they need whenever they need. And they consume data and information at a rate that humans just cannot consume. So you could give them your credit card details or provide your subscription API keys.

Oh, sorry. Well, I'll

just give this a moment.

Well, I think we tried it out earlier and it worked. So maybe let's give it one more shot. says mirroring. Let's try again.

They good.

Just the regular AV issues.

Oh. Let's see if we can start over.

Do you want me to do a window? Will that work better? No, nothing shows.

Yeah, it did. It was not changing slides. So then fun stuff. It's okay. You can just start over.

It's okay. Usual issues. We have a bigger audience.

All right. Well, I'm glad this didn't happen before the live demo. So yeah, let's get started again. Um, as I was saying, um, yeah, with the advent of autonomous agents, there is this need for a new payment infrastructure to to sort of pay for to help and empower these AI agents pay for the content and the data that they need and at this frequency and scale at which they operate. So why microp payments?

Then let's think of a typical autonomous agent. Consider you're interacting with a deep research agent. Um you give the agent your query. The agent's first step is probably scraping through a few thousand URLs. Some of those may be behind a payw wall.

Um, and you probably have subscriptions for a few of those like I don't know like a New York Times or a Wall Street Journal subscription, but then the agent needs to access some content which is which for which you do not have a sub subscription for and there is no way that you can preemptively guess what data your agent needs um in in generic cases like these. So what what do you do in that case? If if you do not provide uh the ability for the agent to that access that payw wall data, the quality of your research is going to be subpar. And let's think of it from the from the point of view of the content creator and the content provider as well, right? They built this payw wall to cater for humans.

uh and now there is this increasing customer base of AI agents uh who are trying to access this data that they have and there's no way for them to monetize it. They're losing out on this potential customer if they keep paywalling and restricting access. Now how do you solve for this? That's where microp payments come into the picture because these agents they want to access a particular article or a particular subset of data uh and the content provider can make it available for them at a fraction of a cost of a subscription. But and that's where the microp payments come into the picture.

Now you ask me how can this interaction work? Well, turns out um it's always been there. Uh in in the original HTTP spec, you had an error code called 402 payment acquired. U the original authors always had this intention and or they had envisioned that payments would flow over HTTP someday, but the rails just weren't there. Now with crypto and blockchain, it's possible.

So, enter X42. So, as long as you have an agent that is a that has a crypto wallet and the merchant or service provider can process or support blockchain payments, you you solve for this payw wall access problem. So, no need for uh API keys or no need for the agents to sign up as they go. They see a payw wall content. They see what the payment requirements are.

Sign a transaction. Request it with the pay uh cryptographic header again. And then you get the data. And this is no futuristic vision. Uh as I'm sure some of you know, um right now this is the reality.

Currently 161 million transactions have been processed with a net volume of $44 million. And this is just over the last six months and it's growing. So what's the challenge of scaling this? Well, if for for one, the problem is bottleneck of gas. We talk about um these microp payments and subs and denomination transactions.

Blockchains and stable coins in theory actually can enable even lower denomination transactions. So you go from the 1 cent to $1 transactions all the way down to nano payment territory which is one microcent. But the problem is gas fees. when you try to when you try to like u settle a one microcent transaction on chain, you're paying a gas fee in thousands of magnitude higher uh compared to the actual transaction value. And if you start scaling this uh over hundreds and thousands of transactions, you're just paying this it the economics just don't work.

So let's look at some numbers. These are like average transaction fees. And if you see at 1 cent level, the fees are still a significant chunk of or of significant value compared to the actual transaction amount. And it only gets worse as the number as you load the denomination of the transaction. But network fees and gas fees are important because they prevent uh abuse and spam.

And then there's also this other problem of network congestion. the scale at which these AI agents operate is is so uh so large that you can clog the network very easily and we want to come up with we wanted to come up with a solution that solves for all of these challenges. So what we've done is our solution separates onchain finality with off-chain payment intent. And what I mean by that is we can see that from from the video that we uh from a live demo that we're showing. So I'll pass it on to Bush for that.

Thank you for the setup Hershel. So let's take a look at the live demo. The fun stuff. Um so here let me walk you through what you're seeing. So on the left side we have set up a bunch of agents.

All of them are on different chains. Um they have different wallets. Now each of these chains have different finality and different gas requirements. Uh on the right side you see merchants. All merchants are accepting payments on all of these different chains.

Well, a merchant needs to accept payments on all of these different chains because you don't want to potentially lose out a customer just because you don't support a chain. Like agent can be on a chain which you don't support and that's a potential lost customer for you. Um we make this integration very easy um with like just a few lines of code without having to deal with all the onchain integrations and we'll dive into that later. Let's start some microtransactions between both of these parties. So as you can see these are live microtransactions happening right now between agents and merchants um randomly uh between any agent and any merchant.

The thing which I want you to see is on the left side on the top here you have gas paid and in this session we have seen what like 8 500 600 transactions. The gas pays paid is absolutely zero and we are operating at a very small payment amount here which is 0 01 which is what we call microcent. The gas savings the number of gas saving is much higher than the actual value transferred in the session. And let's look at it in percentage form. That's like hundreds of thousands of uh magnitude of the payment value itself.

Well, to be fair, the value we are transferring is very small. So, let's change it. Let's go to um one/10enth of a scent. And you will see that even at one/10enth of a scent, it's a meaningful percentage of gas savings. Uh and even at 1 cent, you will see this uh scale the same way.

like it's going to it's going to save you gas. So let's stop this for a second. Gas is only one part of the story. By using this kind of a solution, you don't need to worry about gas and enable like nano payments from microp payments level. Uh on the right side you will see these are the responses from our settle endpoint uh in X42.

One thing to note is all of these are happening on different chains but the latency is consistent across the board. So regardless on whichever chain you're using, you can see that the latency stays the same. And this is great for the sellers because now they don't need to wait for onchain finality and they can just um reliably get a response in a few hundred milliseconds and and unlock whatever the merchant was buying. And overall for both buyers and sellers, you can create a consistent user experience uh because you can be guaranteed to get a response in a few hundred milliseconds regardless of the chain of your choice. And lastly, I think now the merchant got paid, right?

It got money across all of these different chains. Great, we got some money. Uh the problem is fragmented liquidity. Now you have USDC spread across what different chains. Now you need this money to do different things.

Um you want to consolidate it many times and often that requires like bridging uh rebalancing etc. But that's where circle gateway comes in. Both these merchants and um agents are set up with gateway wallets. So then you can consolidate liquidity across chain. Let's take an example here.

I'm going to move liquidity from base and Avalanche. Let's say I'm going to move five bucks and consolidate it onto our test net. And this this is possible for like 13 chains which gateway supports including Solana. I'm going to open up the mint transaction. You can see it was confirmed within less than 500 milliseconds.

We've minted 5 USDC from two different chains to my merchant wallet. And that's it for the demo. All right. To recap the features of our solution, um yeah, so what we have here is transaction sizes. So our solution supports transaction sizes of um value be as high as your blockchain supports and to as low as one microcent right deep into the nano payments territory right um which was not possible because gas fees are very high then you don't pay gas so it's gas free so both the buyer and the seller do not pay gas at a per transaction Third very an important feature is you you get instant liquidity across chains.

Um for merchants you you can easily manage your capital across different chains and avoid the problem of fragmented liquidity. And finally it's X42 compatible. So the demo that you saw the interactions between the agents and the merchants was happening using the X42 workflow that most of you are familiar with. Now let's take a look at how it actually works. So we have this hybrid architecture where we do offchain aggregation of these transactions and we decouple it from the onchain finality.

Um, now to take a deeper dive, what ends up happening is the agent initially deposits some money into the gateway contract. Think of it as your spend limit, right? It could be per session, per for 7 days, for every month, what have you. Then it interacts with a seller in the regular X42 form. And once the seller receives a payment intent from the buyer saying, "Hey, I want to access this resource for 1 cent and they sign that EIP30009 authorization."

The seller can either directly call our APIs or through their facilitator. Here's where the key component is. Once the once we receive this EIP 30009 authorization, we verify the signatures. We verify whether the buyer has enough balance to support the transaction and immediately deduct that amount from the off-chain balance of the buyer. This prevents double spend and we return a success in about 200 to 500 milliseconds.

After that the seller is free to to provide whatever the re whatever resource the buyer requested and that's because circle guarantees at that point that the seller will be paid the money that the buyer promised and then on the circle side we aggregate all of these batches uh batch transactions uh into a single batch inside a trusted execution environment and broadcast one single transaction on chain periodically. This also prevents network logging and it helps prorate the gas cost across thousands of these transactions. Next, this is your regular X42 flow. So, you have your buyer, you have your seller, the facilitator, and the blockchain. And let's move on to the next slide.

And this is where our solution fits in to the X42 flow. uh you the use case is the same. The buyer is trying to access a resource uh that is paywold from the seller and now the first step is yeah uh you have you deposit into the gateway contract and then the buyer requests a resource from the seller. Upon receiving the payment intent from the buyer, the seller calls the facilitator or our APIs directly and we verify whether the payment is uh valid or not. Note that they no longer have to call the verify endpoint.

The settle endpoint is all you need to do because our settle endpoint is so fast that once we verify um whether the authorization is valid or not the seller is free to return the resource to the buyer and from there circle basically aggregates all these uh transactions and then aggregates it like I mentioned earlier broadcast it on chain and once this transaction is confirmed on chain in the seller is free to use the money that they were promised and and then consolidated across chains like Bushut uh showed in the demo and that's the flow. So what's next and who is this for? So we build a solution for a variety of audiences like one is facilitators. Facilitators already add lot of value added services in the X42 ecosystem. This is another feature they could add just just by within a few lines of code.

They can offer gas-free payments to all the people who use their facilitator. Additionally, like we spoke throughout the session, it's for agents who don't want to spend gas and can instantly settle across chains. Well, for service providers, it's an interesting one because now you can monetize your services up to one micro scent and and that enables new business models that you can come up with. uh for chains it's an important one because you want to serve these new customers which are AI agents but you don't want to clog your network so this allows you to bring gas-free nanop payments to your own chain and lastly this is something we learned through discovery um and like talking to a bunch of folks I think right now we have been speaking through an agentic lens all throughout the talk but through by talking to developers in the ecosystem we learned many people are creating apps for price sensitive audiences and they want to use this beyond on agentic use cases in their own apps. Um it's available right now on 12 testn net chains.

We are always growing USDC and our solutions on multiple different chains. Um you can get early access. So you can scan this QR code and fill out a very small form. Um just describe your use case and within a few days our team will reach out to you with early access. Uh thank you.

Automatic transcript — names and jargon may be misspelled.