WEBVTT NOTE Sentence-level transcript of https://www.youtube.com/watch?v=Fu45geO3zX8 NOTE One cue per sentence. Cue ids are the line anchors on /transcripts/Fu45geO3zX8.html. A cue ends where the next begins, or 2 s after its last word. s1 00:00:00.920 --> 00:00:02.120 Hello everybody. s2 00:00:02.120 --> 00:00:03.280 I'm Armandus. s3 00:00:03.280 --> 00:00:07.680 I work at Alifeia Bio where we focus on life sciences topics. s4 00:00:07.680 --> 00:00:13.240 More specifically, we work on immuno-oncology and immunopeptidomics. s5 00:00:13.560 --> 00:00:23.880 While we work on science topics, we use ourselves and provide our clients agentic automation or agent AI solutions. s6 00:00:23.880 --> 00:00:34.160 However, we found a certain gap which we would like to discuss today and we have a solution which we're working on which we would like to share. s7 00:00:34.160 --> 00:00:36.680 It's called Froglet. s8 00:00:37.520 --> 00:00:49.360 But to start with, I would like to ask what do you think is most valuable work that agentic automation can do? s9 00:00:49.880 --> 00:00:56.680 My guess, uh scientific research automation would be quite high up that list. s10 00:00:56.680 --> 00:01:03.520 However, from my experience, adding more tools to agents will not suffice. s11 00:01:03.640 --> 00:01:13.920 Because science and especially lives in in specially in life sciences is inherently collaborative process. s12 00:01:14.360 --> 00:01:24.640 And in order to enable agents to collaborate, we need a way to have a verifiable chain of receipts. s13 00:01:24.640 --> 00:01:25.800 What do I mean? s14 00:01:25.800 --> 00:01:39.960 We need a solution which can provide these receipts proving every step, ensuring that every result can be trusted, and enabling repeatability as well as collaboration at scale. s15 00:01:41.520 --> 00:01:45.720 Having these pieces in mind, let's go to the next step. s16 00:01:45.720 --> 00:01:54.600 So, stepping back, um we imagine agents as cooks in the kitchen. s17 00:01:54.600 --> 00:02:03.120 Giving them more tools or better tools, it improves kitchen's efficiency. s18 00:02:03.200 --> 00:02:10.399 Better knives, more pans, more ovens definitely boost the speed and quality. s19 00:02:10.440 --> 00:02:14.600 But it only enhances the local work. s20 00:02:14.760 --> 00:02:21.120 On the other hand, scientific work is not cooking alone in your own kitchen. s21 00:02:21.120 --> 00:02:26.520 It is closer to running a Michelin-star restaurant. s22 00:02:26.560 --> 00:02:45.880 To overcome this the the outcome depends on providers and quality of your produce, high-level services, as well as an an ability to consistently deliver the same quality dish again and again and again. s23 00:02:45.880 --> 00:02:50.040 You cannot bring everything into one kitchen. s24 00:02:50.040 --> 00:02:51.760 The challenge isn't local. s25 00:02:51.760 --> 00:02:53.520 It's not local tools. s26 00:02:53.520 --> 00:03:01.160 The challenge is aligning the entire supply chain that it's repeatable and consistent. s27 00:03:02.840 --> 00:03:12.680 Today, we have already agents with plenty of tools and increasing in volume, number, and quality. s28 00:03:12.680 --> 00:03:25.400 And we also have on the hand data and specialized analytics algorithms and compute which is distributed across different organizations and siloed. s29 00:03:26.600 --> 00:03:43.440 So, our vision is that whenever AI agent workflow automation matures, organizations will give to the agents not only tools, they will give them budgets. s30 00:03:43.600 --> 00:03:54.480 We're already kind of doing this in a primitive form by allocating tokens for token budgets for a specific task. s31 00:03:54.480 --> 00:03:57.360 But, what I'm talking is a bit broader. s32 00:03:57.360 --> 00:04:01.920 It's allowing agents to manage their own budgets to achieve their own goals. s33 00:04:01.920 --> 00:04:13.360 Whether it's to discover services, request data, negotiate execution terms, or pay for work across organizational boundaries. s34 00:04:13.360 --> 00:04:19.120 At that point, an agent is no longer just a cook with a better knife. s35 00:04:19.120 --> 00:04:23.400 It starts to act like a chef. s36 00:04:24.120 --> 00:04:38.560 Uh like a executive chef, uh finding suppliers, ordering ingredients, coordinating work at at the kitchen, and finally keeping the record of what has happened. s37 00:04:38.800 --> 00:04:41.120 That is why we're building Froglet. s38 00:04:41.120 --> 00:04:53.360 An open-source protocol for agents to discover, transact with, and receive verifiable receipt from external data uh data and service providers. s39 00:04:56.680 --> 00:05:09.520 For more detailed information, obviously, go to our page froglet.dev, uh where you will find um general descriptions on how it works, what it integrates with. s40 00:05:09.520 --> 00:05:14.080 It Froglet is designed to be in between the moving parts. s41 00:05:14.080 --> 00:05:17.040 It's not designed to replace, it's designed to integrate. s42 00:05:17.040 --> 00:05:27.160 Integrates with different payment rails, it integrates with different agent harnesses, execution environments, and even transport protocols. s43 00:05:27.520 --> 00:05:33.280 And it doesn't require that every node in the network uses the same software stack. s44 00:05:33.280 --> 00:05:33.960 That's the beauty. s45 00:05:33.960 --> 00:05:38.600 We are not forcing everyone to use the same to to work the same way. s46 00:05:38.600 --> 00:05:43.280 We're just asking everybody to use the same interface. s47 00:05:43.960 --> 00:05:53.120 In the web page, you will find how to run froglet with just how to install it locally just with one command. s48 00:05:53.120 --> 00:05:58.360 Or even how to run it remotely just with one prompt. s49 00:06:00.560 --> 00:06:14.840 So in essence, what we are saying is closed source and the collaboration often turns into a bespoke project, enterprise project that can take years and cost millions before any reusable s50 00:06:14.840 --> 00:06:17.200 workflow exists. s51 00:06:18.040 --> 00:06:24.200 On the other hand, froglet's mission is to make transaction layer much lighter. s52 00:06:24.200 --> 00:06:41.560 Once an organization decides that the data or resource or service is deemed shareable and it's exposed on froglet, an agent can discover it, understand the terms, request the work, and receive s53 00:06:41.560 --> 00:06:44.960 a verifiable receipt. s54 00:06:46.200 --> 00:06:53.240 That setup costs few thousand tokens and takes minutes. s55 00:06:55.440 --> 00:07:03.440 Now, if I may, let me open our website for quick second. s56 00:07:03.800 --> 00:07:10.640 So if you land our if you go and land our homepage, you will see a couple of things. s57 00:07:10.640 --> 00:07:18.080 First, a lot of helper information which allows you or your agent to read through and understand how it works. s58 00:07:18.080 --> 00:07:23.520 And most of the webpage is focused on agent use. s59 00:07:23.520 --> 00:07:27.280 But also we have some helpful documentation for you. s60 00:07:27.280 --> 00:07:31.160 Let me go through a couple of steps in the demo. s61 00:07:32.520 --> 00:07:41.280 So first of all, we define that froglet node by itself is not different implementations at each site. s62 00:07:41.280 --> 00:07:44.280 It is actually exactly the same node. s63 00:07:44.280 --> 00:07:50.440 It just plays different roles, whether it's a provider, requester, or marketplace. s64 00:07:50.440 --> 00:07:57.680 Marketplace itself is just a Froglet node, which is providing certain services. s65 00:07:59.040 --> 00:08:00.920 And what does it help to do? s66 00:08:00.920 --> 00:08:09.240 It It helps to find, to trust, to pay, and to prove that the work has happened. s67 00:08:09.240 --> 00:08:10.120 That is it. s68 00:08:10.120 --> 00:08:16.600 We're not trying to replace any of functionalities of other tools or protocols. s69 00:08:18.120 --> 00:08:27.680 Whenever you are generating a or creating a new node, it generates a key pair for identity and signing the artifacts. s70 00:08:27.680 --> 00:08:33.000 And during the execution, everything is being signed with that signature. s71 00:08:33.000 --> 00:08:42.479 Meaning, everything is being signed in a chain, and the chain is only valid if all data points are not tampered with. s72 00:08:45.000 --> 00:08:58.880 And in in terms of discovery, request uh providers are publishing what services they uh they uh they're providing to the marketplace using the same protocol. s73 00:08:58.880 --> 00:09:08.040 And requester requests a known marketplace just to provide lists of services that are available on that marketplace. s74 00:09:08.760 --> 00:09:17.400 From that that moment on, once the requester identifies the service that they want to work with, the communication is direct. s75 00:09:17.400 --> 00:09:20.360 There is no uh third party. s76 00:09:20.360 --> 00:09:28.680 And everything from quote to deal to execution and receipt happens in one interaction. s77 00:09:30.920 --> 00:09:34.000 Let me jump forward a bit. s78 00:09:34.360 --> 00:09:39.720 So So far, what I showed you is just a node talking to a node. s79 00:09:39.720 --> 00:09:47.400 Where agent comes in uh are the plugins that we placed inside the Froglet core. s80 00:09:47.400 --> 00:10:04.360 It integrates with different execution environments and different harnesses, whether it's Open Claw, Nemo Claw, it's acts as an MCP server or plugin, and it's primarily designed that it is s81 00:10:04.360 --> 00:10:08.280 used by an an agent. s82 00:10:08.440 --> 00:10:12.960 The primary interface for humans should be an LLM. s83 00:10:12.960 --> 00:10:20.520 An LLM should drive usage of Froglet, whether you are a provider or a requester. s84 00:10:38.800 --> 00:10:43.800 So now maybe let's jump to a terminal and I will show you a couple of things. s85 00:10:43.920 --> 00:10:55.400 First of all, I will show you how to how you can use Froglet remotely, meaning that there is a Froglet node running in a remote server where you can access and get a 15 minutes s86 00:10:55.400 --> 00:10:59.360 trial identity which you can play with. s87 00:10:59.360 --> 00:11:04.480 Second, how to run a Froglet locally on my machine here on a Docker. s88 00:11:04.480 --> 00:11:14.760 And third, how I will ask my Claude to interact with local installation and actually configure a service and try it out. s89 00:11:14.920 --> 00:11:27.120 So first of all, what we what I would like to do is to run a command to check whether Froglet try.froglet.com.dev actually exists and it works. s90 00:11:27.120 --> 00:11:28.960 That's fine. s91 00:11:28.960 --> 00:11:35.840 So now what I can do is to create myself a token. s92 00:11:35.840 --> 00:11:40.840 And this this creates a temporary token for 15 minutes. s93 00:11:40.840 --> 00:11:45.520 And you can see it here created. s94 00:11:47.000 --> 00:11:56.680 And then what I can do also in exactly same manner, I can have access to the specific provider token. s95 00:11:56.680 --> 00:12:01.240 And this is a specific provider which adds two numbers. s96 00:12:01.240 --> 00:12:05.480 It's very simple example service provider. s97 00:12:05.480 --> 00:12:08.040 As you see that the tokens are different. s98 00:12:08.040 --> 00:12:12.080 That means that froglet nodes are different. s99 00:12:12.080 --> 00:12:21.000 Now, in order to execute this service, what I need, I need to create a payload. s100 00:12:22.160 --> 00:12:37.720 So, I created payload which includes many things including schema, what kind of service I'm targeting, also which provider, and finally what input I'm providing it to. s101 00:12:37.720 --> 00:12:53.680 So, I'm providing it seven and [clears throat] five, 7 + 5. Now, what I need is to provide this payload to my remote remote froglet. s102 00:12:54.360 --> 00:12:59.520 And this executes, I need to get a deal ID. s103 00:12:59.520 --> 00:13:04.400 And this is where I have a receipt of actually deal happening. s104 00:13:04.400 --> 00:13:11.249 Now, deal has an ID already, but it might have happened s105 00:13:11.249 --> 00:13:11.440 [snorts] s106 00:13:11.440 --> 00:13:13.280 that it takes some time to so execute. s107 00:13:13.280 --> 00:13:25.280 So, I take a for command for loop to actually wait for it, but it depends on the load on that temporary small node that uh that runs on on the cloud. s108 00:13:25.280 --> 00:13:26.640 And here we have an answer. s109 00:13:26.640 --> 00:13:27.360 It's 12. s110 00:13:27.360 --> 00:13:30.280 Fantastic, we have a correct answer. s111 00:13:30.280 --> 00:13:34.880 Um, so we have a entire loop finalized. s112 00:13:34.880 --> 00:13:42.640 Now, in order to run it locally, um what I need is to install it here. s113 00:13:42.640 --> 00:13:46.200 So, first of all, I will clear everything. s114 00:13:46.200 --> 00:14:01.960 As you see, the uh the the directory is empty, and what I do is I what I run is HBS Froglet Dev Agent, and what it will do, it will download and install agent s115 00:14:01.960 --> 00:14:04.400 on a Docker. s116 00:14:04.400 --> 00:14:05.920 Let's wait a second. s117 00:14:05.920 --> 00:14:14.880 It's actually already there, so it just confirms that it doesn't need to download anything and it's working and it's running. s118 00:14:15.640 --> 00:14:29.520 And now I can see that uh there are two Froglets running, um but if I look at at the actually not IP addresses and ports, but I actually look at the node ID, s119 00:14:29.520 --> 00:14:32.640 I see that this is one and the same ID. s120 00:14:32.640 --> 00:14:40.440 And that means that each each Froglet can assume different roles, and at the same time can assume multiple roles. s121 00:14:40.440 --> 00:14:45.160 And we have here two roles as a provider and consumer at the same time. s122 00:14:45.160 --> 00:14:57.040 And just to prove that I'm running it on a Docker, here it is, running on the same ports, same IP addresses, my local ins installations. s123 00:14:57.240 --> 00:15:11.240 So, now Claude, if I open my Claude, first of all, what I need to do uh is to see where Froglet MCP is enabled, and it is. s124 00:15:11.240 --> 00:15:13.320 That's fantastic. s125 00:15:13.320 --> 00:15:24.040 And what I will ask it to do is, first of all, I will change the permissions that it doesn't bug me. s126 00:15:24.280 --> 00:15:35.240 And what I will ask it is use from from that um CP only. s127 00:15:35.360 --> 00:15:39.520 Do not install anything. s128 00:15:42.000 --> 00:15:47.480 And um let's be very specific. s129 00:15:48.000 --> 00:15:51.680 Let's see if it understands constraints. s130 00:15:52.000 --> 00:15:56.640 And I have here prepared a couple of steps. s131 00:15:56.640 --> 00:16:00.520 I won't do it them one by one. s132 00:16:00.520 --> 00:16:05.480 I'll actually ask it to do it all at once. s133 00:16:07.160 --> 00:16:11.520 In steps and provide the specific output. s134 00:16:15.760 --> 00:16:21.560 Great, it even goes ahead and um uh checks what is running. s135 00:16:21.560 --> 00:16:31.240 And what I'm asking it to do is first of all to see what was the status is life, which already did above. s136 00:16:31.240 --> 00:16:42.161 Then if provider and runtime is healthy, then I want it to publish an artifact from a template, which is exactly the same template add two numbers. s137 00:16:42.161 --> 00:16:42.760 [snorts] s138 00:16:42.760 --> 00:16:51.240 Then uh what I want it that it actually finds and calls that service and invokes it locally. s139 00:16:51.240 --> 00:16:56.640 Uh asking it to add again five and seven. s140 00:16:56.640 --> 00:17:02.160 And it can wait a bit because it's an asking for a call. s141 00:17:02.160 --> 00:17:05.959 And then show the output what it got. s142 00:17:05.959 --> 00:17:17.199 So if if we go if we go down here, here is what we see that it actually executes all commands. s143 00:17:19.000 --> 00:17:24.480 And what we see that is actually um done exactly as we asked. s144 00:17:24.480 --> 00:17:27.760 We It has a result some 12. s145 00:17:27.760 --> 00:17:31.040 And it It can create anything that you want. s146 00:17:31.040 --> 00:17:46.240 It can connect to You can ask load to connect to your database and provide row by row access for for uh for a payment or you can have a just your GPU running in the background and s147 00:17:46.240 --> 00:17:50.480 welcome the requests for computa- any kind of computation. s148 00:17:51.080 --> 00:18:04.040 In essence, it's a very simple way how to use it, but it packages a lot of um underlying protocols and underlying tools where now you're not shoving everything into s149 00:18:04.040 --> 00:18:06.480 a uh LLM context. s150 00:18:06.480 --> 00:18:10.640 It just has a services that it needs to interact. s151 00:18:10.640 --> 00:18:17.160 And underneath there is a receipt, there is a payment, there is a execution environment. s152 00:18:17.160 --> 00:18:25.560 And uh there is a negotiation and marketplace uh hidden and it's uh doesn't stuff your LLM with context. s153 00:18:26.240 --> 00:18:30.520 So, going back to the slides. s154 00:18:31.560 --> 00:18:34.360 Some reason it scrolled down. s155 00:18:34.360 --> 00:18:36.880 Going back to the slides. s156 00:18:37.280 --> 00:18:42.120 Um So, what we what we have uh on our hands. s157 00:18:42.120 --> 00:18:47.760 Uh we have an open protocol called froglet which um has multiple things. s158 00:18:47.760 --> 00:18:59.640 First of all, uh it uh provides a way uh to to find resources outside the organizational boundaries. s159 00:18:59.640 --> 00:19:06.280 Then it allows your agents to actually execute uh remote uh commands. s160 00:19:06.280 --> 00:19:14.440 And third, uh what it allows, it allows to have a verifiable receipt of receipts of what has happened. s161 00:19:14.440 --> 00:19:19.360 And I think that actually enables a collaborative science. s162 00:19:19.360 --> 00:19:26.280 That enables a progress of a perpetual science automation. s163 00:19:26.280 --> 00:19:31.280 So, please join us either via Froglet or Althea Bio. s164 00:19:31.280 --> 00:19:33.320 And please join us on GitHub. s165 00:19:33.320 --> 00:19:36.080 And thank you for listening.