Hi everyone. Uh so, my name is uh Blessing Adjele. I'm with the uh Arc team. And for those who are not familiar, Arc is uh the Arc is a cool L1. And today, I'm going to pretty much walk you uh bounty and show you a little demo of what we're expecting from you.
Uh it's actually a new sample apps that we are about to release. We'll be releasing it after this uh workshop. So, I'll post the link in the Arc uh Discord channel or like the E Global Discord channel. So, you don't have to worry about that. Uh so, quickly, let me just take cuz I I understand I have about 15 minutes, right?
So, let me just take a few minutes to walk you through the bounties so you know uh what we expect from you, what we expect from the teams, yeah. So, the first use case that I see here is the best smart contract on Arc with the advanced stablecoin logic. So, that means any sort of like smart contract uh that is deployed on Arc and has like really solid stablecoin logic. That could be range from escrow to like lending and so many other use cases. Uh and also payroll as well leveraging USDC and ERC.
And also, the next one is uh abstracted USDC apps using Arc as a liquidity layer. You can use Arc as the liquidity layer and that allows you tap into other all all other all other primitives that we support like gateway, CCTP. And you can use Arc as the settlement layer cuz of course, gas is in USDC and it's fast uh finality. So, uh best agentic economy with nano payment. Uh so, as a matter of fact, uh the demo I'll be showing you today is actually agentic use case.
So, I'll give you all like the baseline how to think about where what to think about when you're building any agentic use cases and how you can use our latest and greatest nano payment. We actually just released it uh last month and I personally love this product cuz it solves a lot of challenges when it comes to paying in substance of the dollar and gas cost that is attached to that. So, uh then the next one is best prediction market built on Arc with real world settlement. So, that's that's just another use case that they were expecting. So, without wasting so much time, let me actually just jump into the into the into the demo.
So, uh this is the sample apps that I told you that we'll be releasing. Uh so, we'll get the link to you guys on Discord. So, don't worry about it. But, um on a high level, the sample application actually just Sorry, I can hear myself back. But, uh on a high level, the sample application lets you interact with X402.
And how it works is that you use X402 to make some type of payment with agent. And uh if you're not familiar with X402, X402 allows you tap into the HTTP protocol 402. And with that protocol, you can uh you can a buyer can make a request to a seller and then the seller can say, "Okay, before I grant you this request, you need to fulfill a payment requirement." So, say for example, I'm the buyer and you guys are the seller and I say, "Oh, I I need access to the content behind your paywall." You can say, "Blessing, before you access the content behind this paywall, you need to pay maybe USDC, maybe 0.
001 USDC." And when I pay that, you then grant me access to the content. So, think of it like this, the interaction has to be uh both ways. So, I send a request to you and then you eat me back with the 402 uh a response. And with that response, I can take that and then retry that request again and then I make the payment.
So, that's the entire premise that nano payment combined with our protocol, which is known as gateway, is built upon. And I'll show you how the sample application works, right? So, you can see right here the paywalled uh endpoint. So, we have a couple of endpoints in this sample application where uh you can receive a quote, which costs 0.001 dollar and that returns a premium inspirational quote, just for example, right?
Then you can also uh request for data set, right? And that costs 0.01. And that returns a small JSON data set, right? You can receive compute, right?
And this is when you let the AI agent do some sort of like test analysis on a submitted content. And then you can do agentic task and that costs 0.03. And that returns a a a closed step for a treasure hunt in this particular example. So, this paywalled endpoint uh something needs to be something is happening on the other end.
And for the agent for the agent to complete that, they need to get back that 402 response and then they need to make the payment. Don't worry, I'll show you how how that that works right now. So, I actually put together this uh diagram right here to actually explain the flow so it begins to make sense to you. So, the way the sample application actually works is that uh the buyer uh with a particular address funds the wallet and then funds the wallet for example like 5 USDC, right? And the gas is like 0.
01 because we're batching the entire transaction, which happens off-chain and that's the way nano payment works. And when uh and when the when the buyer's address has been funded an ephemeral wallet agent is being created, which is a temporary wallet. It's created for the agent. And that's the sample application I'm working working you guys through. And uh what happened is that they they make a request to one of the endpoints that I showed you.
In this case, they're making a response they're making a request to the quotes uh endpoint to say, "Okay, I'm an AI agent and I need to access the quote that is behind the paywall." And once they send that request, take note, with no payment, right? I you the seller, let's say you guys are the seller, you then eat back and say, "Okay, payment is required with the 402 response." And then this happens off-chain where you sign a payment off-chain signature. And once that's signed, uh the buyer the seller verifies of it verifies that the signature is okay.
We then eat you back with a 200 response saying, "Take the content that is behind the paywall." So, take note, it's very important that you understand what's happening right here that the the buyer is actually not making any payment in this case. The buyer is only signing. And the reason why the buyer is signing, I will explain to you, is because at the end of the day or at the end of a couple of hours, which is the buy the agent rather, at the end of hours, we actually batch up all of those transactions uh that we get from this signature, we will batch them up and then we settle to the seller. And that way, the gas is like very very small because we're batching up hundreds of a thousand uh transactions all into one and then we settle it into the gateway uh gateway balance of the seller.
And I'll show you how how all of that works. So, I I have a couple of key points that you should take note right here that the agent generates a new wallet each each time it runs. So, it's a temporary wallet. It's called the ephemeral wallet. Only two transactions happen up on on chain.
Deposits to gateway so that with that particular balance, the agent can run some use cases. And then the batch settlement happens pretty much later of all of the entire transactions. Zero gas for agentic payment cuz we're batching up the off-chain signature. So, so it's not like we're doing one one transaction anytime the agent runs a particular task. So, and these are the endpoints that we interact.
So, to make this make sense, this is a little dashboard of the sample application. And you can see example of Let me actually zoom in so you can see properly. And you can see example of each agentic request. So, the agent tried to get a request to the data set endpoint, to the quotes endpoint, do some agentic task to get the treasure thing I talked about, a compute task to actually analyze a particular task text. And all of these actually happened in this view.
So, let me run Let me run it so you can see. So, in my terminal, I'm just going to run So, in my terminal, I'm going to run uh okay, let me actually start this again. NPM run dev. Uh So, in my terminal, okay, this is already running. Pretty much.
Okay, that's fine. Uh so, in my terminal, I'm going to run uh NPM run agent and I'm going to do 1 USDC. So, I'm saying that the agent has ability to just spend only 1 USDC. And you're going to see a bunch of transactions actually happen by the agent. You can see that I assigned a spending limit of 1 USDC to the agent.
The ephemeral wallet, which is the temporary wallet that I talked about, has been created. And uh I'm depositing 5 USDC into the gateway wallet. So when the 5 USDC is there in the gateway wallet, the that can be used as like gas gas right for the entire transactions that I talked about that is being happening off-chain. So let's say I'm doing 100 transactions. So the gas will be of those 100 transactions which is like 0.
01 dollars will be taken away from the $5. So these are all of the tasks that the agent is completing. You can see it makes a request to get the data set, it gets back the compute, and you can see all of that. So and in this particular dashboard, you can see how everything is happening. The agent is actually doing task with sub with sub cents of a dollar.
You can see how everything is happening. Before I came here, I did some transaction, but you can see the total payment here is like 100 and like 191. And you can see the amount being spent. And you can see the task that the agent is actually completing. It's getting It's trying to It's making access to get the data set, code, compute, all of that stuff.
And you can see live feed of how all of that is happening. And pay attention to this. You can see that this happened in less than 500 milliseconds, meaning it's like super fast for each of this transaction. And the beauty of nano payment is that for each of these transactions that happened, I'm not paying gas. Right?
They are off-chain signatures that the agent is actually signing. And at the end of the day, all those signatures will get batched up together, and gateway automatically, maybe after a couple of hours, will then settle all of those funds into the seller's gateway balance. Does that make sense? Okay. So So you can see in the dashboard that I have right here.
This is red, but yeah, you can see in the dashboard where I have here that these are the transactions that I've done previously before I came here. Right? This is the seller's wallet. Right? So the seller that has access to the data, to the compute that the agent is trying to make access to, this is their wallet.
And at the end of the day, after gateway has done all of those batching and settlement, it will give it will there will be a withdrawing some balance that the seller can actually withdraw. Right? Because the transaction is not happening real time. They are batched up together, and they can be settled on chain much more later, making this almost zero gas fee transaction that is happening. So I have 5 minutes for question.
Let me know if you have any questions. I'll pause. Let me know if you have any questions. Why are Why are all Why are the tasks actually keeps running? Well, I think we're done, right?
So you can see that the agent has successfully spent 1 USDC. Don't forget we assigned a limit of 1 USDC, and this was like $1.5 by the time I started this workshop. And if you take a look at the balance, you can see that it moved from 3.5 to 4.
5 USDC. So the seller has gotten all of the 1 USDC that the agent spent on kind of like 100 task that we did with a sub cents of a dollar. So I think we did about This was This was built By the time I started this workshop, this was like 170 something. So we did I can check right here. I think it's in my terminal.
Yeah, we did 98. Yeah, 98 tasks there. Hi. We were wondering how will the seller know once you pay? So say that again.
So when you pay Yeah. When gateway pays. Yeah. And how does the seller know live instantly that you have paid? Okay.
So if you see the you see this right here. The money is still in like the seller's off-chain balance, right? It's not right in the seller's gateway balance. But by the time it becomes available, the seller will be able to tell right here after a couple of hours because gateway automatically does the settlement. Does that make sense?
Not not really. Let's say the seller wants to instantly know that I've paid. That you've paid. That the agent has paid. Yeah, agent or me, whoever.
So using nano payment, the the seller can actually see these are completed tasks. These are completed payments. But they are signatures, right? That gets batched up, and the real cost actually gets settled at the end of the day. Does that make sense?
So the seller would know after I make a payment that a payment has been done successfully. You can see these are examples of successful payment, and the seller can actually see. So in the case of our example, the seller probably has access to this dashboard, and they can see everything that is happening here. And they can even click it, and they can see them happening on on on on chain, right? On ArcScan.
They can actually on the Arc Explorer, they can actually see. And then they can verify that okay, this this are this are actually happening. Does that make sense? So the seller would have to use Say again. So the seller would have to use Arc No no no.
So So this nano payment supports multiple chain. It's not only Arc, right? Let me get you to the developer documentation to show you all the uh Yeah. Supported network. So you can see the seller can uh use supported gateway chains.
So these are all the chains that the seller can actually All of this can happen on this on all of these chains that I showed you. So you can even give it can give you idea of things to build. You can build a multi-chain stuff where maybe settlement happens on Arc because it's gas gas is in USDC, and it's fast. And then you can wrap balance from like base, Ethereum, Arbitrum, and all of that. Hi.
Hi. I saw that you had an endpoint called compute. Say that again. Compute, yes. Compute, yeah.
So is there a way to pay uh for your service which doesn't have a deterministic value? Let's say a compute cost costed like uh 5 USDC, but it's supposed to cost like uh 50 cents. Yes. So So So the compute endpoint actually interacts to like OpenAI API or like Cloud API, and it's actually doing proper LLM inference task what I've just showed you. Does that make sense?
Oh, sorry. So and that does not deterministic. Does that make sense? Got it. Yeah.
Okay. So you can actually take this demo, and you can plug in your own OpenAI API or your own [clears throat] Cloud API, and have the agent run task using the API. And the compute payment is what is what you're seeing. Got it. Yeah.
Thank you. You're welcome. Yeah. Time. Thank you guys.
Thank you for having me.
Automatic transcript — names and jargon may be misspelled.