Kshipra Dhame I A comprehensive showcase of AI Agents and Agentic LLMs I ETHGlobal Cannes 2025

[Music] Hello everyone, I'm Shepra, a developer advocate at Patchet AI. As a part of my role, my main responsibility is to provide the right set of information and resources to hackers as well as other community members building using the Fed stack. In this workshop today, we will learn how to build AI agents with Fetch AI. So let's get started. First of all, let's look at the agenda of this workshop.

We will start off with what is Fetch AI. Then we will talk about the Fetch AI text tag and the three key components in the Fetch stack which are ASI 1, agent works and u agents. Next we will have a hands-on session where we will build our first agent. Then let's say if you've built agents using different agentic frameworks, we will learn how to get those agents on agent wars. And finally, I will point you to the right set of resources that will help you get started during this hackathon.

So first of all, what is fetchai? Fetch AAI offers a complete ecosystem for building AI agents, searching and discovering these AI agents. And these AI agents are also capable of performing transactions in the form of feet tokens. So all in all it gives you a complete Asian techch economy. The first component of the fetchi ecosystem is ASI1.

So ASI1 is a web3 native LLM. So I know there are a lot of LLMs out there such as shad GPT, Gemini and so on. But we cannot really contribute to the knowledge base of these LLMs. But with ASI1, you can contribute to the knowledge base of the LLM in the form of agents. Meaning ASI1 has access to agents and it can call a particular agent based on a user's query to get a task done for you.

Next, we have agent war. So agent wars is a marketplace for AI agents. So on agent works you can build your agents, you can host your agent, you can also search for agents built by other people. You can use these agents in your application or you can use these agents to get a simple task done for you. Finally, how do you build those agents?

So to build those agents, we have U agents. Using U agents, it is a simple Python package or a Python library. You can build simple agents that can perform some tasks for you. So these agents have different handlers to send messages from one agent to another. You can also send feet tokens like I mentioned from one agent to another.

So we will be covering some bits of it in the following workshop. Now now let's look at the fetch text tag. At the bottom of this text tag is blockchain. So whenever you create an agent using new agents, the agent gets registered on Almanac and every time you send a message from one agent to another or every time you send feet tokens from one agent to another, it gets recorded on blockchain. And since it's on blockchain, these records are immutable, meaning nobody can change these transactions or can uh or can amper records.

Next we have the agentic framework. So we have two Python libraries to build agents. One is U agents and the other one is SDK. So during the hackathon we highly recommend using new agents as it is much simpler and it has a lot more features than the fetchi SDK. The SDK is built on top of few agents and we recommend using SDK for more industrial level applications or solutions.

So we highly recommend during the hackathon you stick to the UA Asians package. Next we have ASI1. Like I mentioned ASI1 is a web3 native LLM which has three different models. It has ASI1 mini, ASI1 extended and ASI1 fast. So you can use ASI1 mini if you have simple tasks that you want to get done.

If you want to have some deeper understanding or a deeper thought process before the LLM answers your question, that's when you can switch to ASI1 extended. And if you want to get something done quickly, then you can use ASI1 fast. This ASI 1 LLM has three main features which is it is an agentic LLM meaning it can search for agents based on the type of query that you're asking and then respond to you if if there's a need for some real-time information or if there's a task that needs to be done that the LLM by itself cannot do. Next, you also can add knowledge graphs to this LLM. I will talk more about knowledge graphs when we get to it.

And on the right here we have agent wars. Like I mentioned, agent wars is a open marketplace for AI agents. But it is also a platform or an IDE for hosting agents. It has the mailbox feature. So we will talk more about the mailbox feature and when to use mailbox and when not to use mailbox during the hackathon.

It also has some monitoring tools. So you can check different information about your agent such as the number of interactions associated with the agent, the ratings associated with the agent and so on. So we will have a look at that as well. Next it has search and discovery. So you can like I said you can search for different agents on agent wars.

You can make use of those agents in your applications or the solutions that you're building. And some of the agents if you build agents with chat protocol you can also chat with that Asian directly in natural language or in human language like we say. Now let's look at the tools available for you to build agents and to deploy them during the hackathon. So first of all to start building your agents you can use new agents and if you want your agent to be smart enough you can also have the ASI1 LLM integrated in it and have the chat protocol if you want to talk with your agent in natural language. Now let's say you have built agents using different frameworks such as lang chain, crew AAI, autogen or MCPs and you want to bring those agents to agent wars.

So for that we have something known as the u agents adapter which will help you connect these agents that are built using different frameworks to agent words. And let's say if you've built some U agent that is on your local machine and you want to connect that U agent to agent words, you can use the mailbox feature. We will be looking at it shortly. Next, you can either deploy these agents on agent work. You can also deploy them on your local machine or you can just spin up a VM and have your agent deployed on the VM so that it's always up and running.

Next to test your agent, you can directly test it on agent words using chat with agent or you can also do it locally through by creating another agent that communicates with your agent just to see if you're getting the right response and so on. Next to discover the agent. So this is the interesting bit here where with ASI1 LLM let's say you type in a query and your agent is good enough to get the task done. For example, let's say I have a cryptocurrency agent and on ASI1 I say please help me convert 50 BTC to USD. Then ASI1 can actually go to agent words and look for the agent that I just built and give us a response back.

So let's look at the first pillar of the fetch stack which is ASI1. Like I said this is an Asentic LLM. So if you see in the UI here as well, we can toggle this the agent switch on and the LLM can then connect with different agents to get to answer the queries. This LLM also has adaptive reasoning meaning it can switch between simple reasoning, short reasoning to complex reasoning very easily based on the nature of the task assigned to it. You can also add knowledge graphs in the form of documents, websites and so on.

We will take a look at that as well. And right now it has four models. ASI1 mini which you can use for simple or basic tasks. It has ASI1 extended. If you require deeper thinking before the LLM answers your question, if it's a complicated task, then you can use ASI1 extended.

You can also use ASI1 fast. It can quickly discover agents or it can answer your queries much more quickly. And lastly we have ASI1 graph. So this is a new model through which uh you can do some data analytics. It will give you some fancy graphs for predicting stocks and so on.

So let's quickly go to ASI1. So we can just say ASI1.ai AI. So you can log in to ASI1 using your Google account or using the fetch wallet. So you can see there are a bunch of features here.

Like I said, there is a agent toggle switch. So this will enable the LLM to search for different agents on agents. We also have web search. So let's say you just require to search for something on the web, you can toggle this button on and it will search for different things on the web. You can also attach different images and files and pass it to the LLM like you would normally do for any other LLM.

And then we have something known as the knowledge graph. So this is the interesting bit here. So to add something to the LLM's knowledge graph. So knowledge graphs what it does is it gives more context to the LLM before it answers your question. If I go to the knowledge graph, you can see a bunch of things that I can add to the LLM's knowledge before I ask any questions to the LLM.

So, I can add different kinds of documents. I've added a few here. I can also provide websites that it should rely on before answering my questions. You can also generate something related to research. You can provide more context.

If you think there is some uh custom knowledge that you have that the LLM should know about, you can add that here. And you can also do some validation. So going back, you can switch between different LLNs such as ASI1 mini, ASI1 extended, ASI1 fast, and ASI1 graph. Let's go back. Okay.

Now, let's say I want to try out some queries. So, first uh let's say I want to check out the latest train and tube status at London White City. So, if I go to the LLM and now if you see the agents toggle is off. I will also switch off the web search so that it does not go to the web to answer my question. And if I hit go, so now the LLM is doing some thinking.

So as you can see, here's the latest information based on the LLM's knowledge. So it's answering me based on the knowledge that it has from 2025. And it's saying that for realtime updates, please go to TfL and get information about the tubes at White City. So this is because at this point the LLM does not have access to agents on Asian ports. Now what if I try the same query again and I toggle the agents switch on.

This time what the LLM should actually do is it should go to agent words and look for agents that can actually answer my query. So like you can see here it's saying normally I wouldn't have access to real-time transport data like brains or tube statuses. So it's going to go to agent wars. I can search the agent wars for for specialized agents and it will now do some uh tailored searches and then try to look for agents that can actually answer our query. Let's give it a few minutes.

Yeah, it has found five top agents. So it has found the transport for London agent. It has found Stratford broadway agent and so on. So if I just click on this agent until the LLM responds back to us. Okay.

So this is the agent which we have for transport for London. It can give us information about line status which means it should be able to answer our query. It can give us information about station arrival, station information. It can help with journey planning, bike points, time awareness, and so on. So, let's go back and see if the agent has responded.

Yeah. So, it has responded. It says, "Here's the latest information on White City Station." So, typically running smooth with trains every 2 to 3 minutes of fe. It has also provided some alternatives such as the Hammersmith and City line is nearby if needed.

You can walk or cycle as well. It has provided some local tips. Check out Westfield London or grab a coffee at the Alice. So this is how the LLM works to answer questions that it might not have it might not have information about. Now let's try another query.

Let's say if I want to get if I want to uh convert 50 FP to USD based on today's value. So if I punch in this query again, the LLM will do some reasoning. Okay, it it says here's the conversion, but it is based on the knowledge that it has until 2025. So let's just open another chat and let's try this query again. So it has understood that the a the user that is me I want to convert 50 ft tokens to USD which is a cryptocurrency to fiat conversion request.

So it goes to agent words. It will again try and select the most appropriate agent for this task and then it will try to answer. So it has found about five agents that can answer this question. So and okay so this is actually an agent that I have built. So we can go to this agent and look at the logs.

So we can see at on 26th June around 1436 it has received a request which is please convert 50 uh tokens to USD. It has just patched the envelope. it. This is the result that it has. Now, let's go back to ASI1 to see if we've got the response.

So, yeah, it's saying here's your conversion. 50 FT is currently worth 1616 USD. So, this is how the ASI1 LLM is different from the other LLMs where it can actually go to agents that are designed for very specified tasks or very complicated tasks and it can get the job done for you. So moving on, next up is Agent Wars. So like I said, Agent Wars is a marketplace for AI agents and we have quite a lot of agents deployed on agent wars.

It also offers you an IDE. It gives you a way to look at the stats and you can also chat with agents directly on agents. So let's go to AsianPourse.ai. Okay.

So I was already here. It's checking if I'm a human. So again on Asian course as well, you can either log in using your Gmail account or you can just log in using the ASI1 wallet. So this is what the Asian course marketplace looks like. You can see we have almost 2.

61 61 million agents deployed and they have exchanged some 130 million messages. So that's really crazy and there are a bunch of agents available here and we will also be looking at so when we will get on to the hands-on session we will look more into the IDE that agent course has. So if you switch to agents you should be able to look for something known as create an agent. So we will get to that shortly which will which actually is the ID for building agents. You can also look at the stats associated with your account.

So you can check how much data transfer out limits you have, what's the expiry, how much storage have you utilized, how much messages uh can you still exchange between agents and so on. And let's go back to the marketplace. Now let's say I want to check the stock price of an agent. So if I search for some agents that will give me stock price. So this is one agent here which can give me the stock price and you can see here that this agent actually has the Asian chat protocol.

So whenever an agent has the Asian chat protocol you will look you will see here that it will have the a chat with Asian button. So with the chat with Asian button you can actually go and send a message in natural language to this agent. So if I just say I want to get the stock price of Apple. So what it will do is so since this is not my agent I won't be able to check the logs but it will send this query to the stock price agent and in natural language and then it will respond us back with the price which is 201.56.

So this is how you can interact with agents that have the Asian chat protocol enabled. And during the hackathon, we highly encourage you build agents with Asian chat protocol so that you can also chat with these agents with natural language and they are also discoverable through ASI1. So for agents to be discoverable discoverable through ASI1, they must have the Asian chat protocol. So this was one bit here but there could be different agents which may not have the agent chat protocol enabled. So like for this one here this is the technical analysis agent.

It does not have the agent chat protocol but the user has actually the developer has actually given a really nice usage example here. All that you will have to do is copy in this code and start interacting with the agent. So during the hackathon you can use some of the agents built by other people and they are a part of the Asian marketplace or you can always create your own agents and then use them in your applications or your solutions that you would be building. So let's go back and next we have uations. So like I mentioned the basis for everything is u agents because it is the framework that will help you build those agents.

So whenever you build an agent it will get registered on almanac. So the minute the agent is registered on almanac it will have an identity or an address through which you can communicate between two agents. So let's say I have one agent and my friend has another agent. These two agents can send messages to each other and communicate with each other. Next we have protocols.

So these are nothing but like I said the Asian chat protocol which is nothing but a set of rules defined for agents to communicate. So there are different types of protocols or you can also create and define your own protocols for agent communication. Next we have mailbox. So let's say there are some agents which you cannot build on agent words but you want these agents to be connected to agent words. That's when you can use mailbox.

We will be doing a hands-on on that. And after that, like I said, these agents have wallets. So they have the fetch wallets and you can transact using those wallets, using the feet tokens. Now here comes the fun part. Now we will build our first new agent on agents.

So let's go back again. So let's go back again to agent tours. I will just quickly open another window. Agents.ai.

Now if we go to agents, we will click on create an agent. And over here you can see some templates already there. So I'm simply going to say I'm going to use your first agent. And I'm going to call this my first agent Shra. So this gives you a sample code with which you can just get started.

So like I said from new agents which is nothing but the new agents package we import two modules which is agent and context. So all that you have to do to initialize your agent on agent words is you say agent equals and then you initialize that agent object. Whereas if you're creating a local agent, you might have to pass in a few core parameters such as main, seed, code and endpoint. So we will take a look at it, don't worry. And then these agents have different handlers.

So or triggers. So it come. So likewise we have handlers for shutdown. We have for sending messages. There is something known as on message that we will just check in a minute.

So when this agent will start all that we want to do is using ctx.logger.info info I want to print hello I'm an agent and my address is so I I mentioned that whenever we create an agent each agent will get associated with an identity or an address so I all that I'm doing is I'm just printing this agent's address so now let's start our agent also you can very easily switch between testn net and mainnet to decide where you want this agent to be deployed you can either deploy it on testn net or you can deploy it on mainet So for now I'm just going to leave it to test net and I will click on start agent. So you can see it has started and it has also printed hello I'm an agent and my addresses. So a typical agent address will start with something like agent one and then a bunch of random characters but these are always unique to each agent.

Then in the overview section, you can update the readme for this agent. So this is very crucial and very important especially if you're building agents and you want those agents to be searchable on the marketplace or to as it is very important that you give a relevant read. So now in this case all that I'm going to say is this is my first agent. But based on the task the agent is doing you can add complicated readmes like we just saw. You can also provide some uses usage examples or some sample queries that people can try using your agent.

Then we have the ratings and and analytics tab. You can check out. So right now since this agent is fairly new and there have been no interactions with this agent. So the interaction count is zero. You can check the ratings associated which is 2.

25. Not too bad. And then it also shows similar agent. So, I'm pretty sure there are a lot of people out there who've just created their first agents and that's why it's saying that my agent is very similar to these agents here. It also shows the alltime searches, the last 24hour searches and so on.

You can also check the top search terms. So, it might not appear just yet because I've just created this agent. So, this is this is about it. This kind of gives you an idea what the Asian vers IDE looks like. Now we will see how to create two agents that can communicate with each other.

So I will go back again to this agents tab and I will click on create an agent and this time I will select the agent communication template. So it will create two agents for me. It will create Alice and Bob and then they will be sending messages to each other. So it will be quite interesting. Click on next.

Click on create. So you can see it has created two agents for me. One is Alice and one is Bob. I will open Alice. We will go to build.

Okay. So in Alice, we first initialize the agent. And now we have something known as a data model. So this is sort of let's say I want to pass a message from one agent to another. So for the other agent to know what is the kind of message that I will be sending we define the data model or the envelope for that agent or that message and what I want this agent to do it on startup.

So there are two handlers but first let's talk about the startup handler and then I will talk about the onssage handler. So on startup I am just saying just about to send a message to Bob and using this line here which is a wave ctx. I'm actually sending a message to Bob. So to send a message to Bob, I would require Bob's address and I would also require the data model that we just defined up here, which is the request data model. And I'm saying hello there Bob.

And finally, I'm going to say message has been sent to Bob. But first, I need to verify if the Bob if the address that we have here for Bob is correct or not. So I will quickly again in a new tab open agents and then look for Bob and we can see it's ending with 4 RRT and let's go back here. So this one is also ending with 4 RRT. That means we've got the right address for Bob.

And once we once we send a message to Bob let's look at the code for Bob. How will how will Bob handle this message? So first and foremost we initialize the agent we define the data model which is the request data model. So Bob knows that it will receive this kind of message from Alice and when Bob receives this message that is on message. So this is the onssage handler and you in the onssage handler you define what is the kind of message that you will get.

So in this case it's request and then it will say using the ctx.logger it will log the message. So it will say received message from sender. So over here it will print the agent's address and then print the actual message and then it will verify if the sender was Alice. So if Alice is if the sender was Alice, it will send back a message which is hello there Alice.

Else it will just say hello there friend. So now if I go back once Bob sends a message it will it will sort of be dealt using the onssage handler that we have here and it will lock the message. So first I'm going to go ahead and start Bob and then I'm going to start agent Alice. So it says just about to send a message to Bob. Message has been sent to Bob.

So let's go to Bob. So it says received the message from and this is the ID. So it ends with GND meaning it has to be Alice's address. So if I go back to overview, so Alice's address does end with G&D. So that means uh agent Bob should actually send back a message saying hello there Alice.

We say envelope dispatched to posted agent which is Alice. Now let's go to Alice. And we can see that it has received a message from Bob saying hello there Alice. And that's how two agents can communicate using ctx. And the onssage handler.

So this is something that you might require during the hackathon and the chat protocol is sort of built on top of this. So this is this will be really helpful for you guys. Now we can actually check the interactions associated with Bob. So we can see the alltime interactions associated with Bob are one because it just received one message from Alice and that's why it says the alltime interactions are one. Now going back.

So there are different approaches that you can take while building your agents. So the first bit here to understand is not all kinds of agents can be hosted on agent wars because not because agent wars does not support all Python packages due to security reasons. So if you have some agents that require a specific a specific package that is not supported by agent vers then we can create local agents and to find out the list of packages that are supported by agent war you can check them you can check them here. So if there are some packages that you need and they are not listed here you can create local agents and then you can connect these local agents to agents list. So we will be checking that out just now.

So you can create a local agent and to connect this local agent to agent worse you can use mailbox. So sometimes let's say you're building some application and you want it to be private you can create local agents and you can uh just keep them local on your machine or your but if you want those agents to be discoverable through Asian ps or through ASI1 then you can connect these local agents using mailbox. So let's go back to Asian words for a minute. Now what I want to do is I want to create this Asian Alice again on my local machine. So what I will do is I will simply copy this code.

Okay. And I will go to VS code. So I have VS code here and I will open my terminal. So I like to create a virtual environment before getting started. So I will just say three-ph MD space MD.

So this will create a virtual environment. It's taking some time. Yeah. And now I will activate this virtual environment. Now let's install U agents.

So on agent work U agents is installed by default but on your machines you will have to do a simple clip install new agents. This might take a few seconds. There we go. It has been installed. I will quickly verify the version of UAE.

So I'll do a quick three show u agents and the latest uh UAEN version is 0.22.5. So let's go ahead and create our local agent. So I'm going to call this file local agent.

py. And now I'm going to go back to agents. Copy. Make sure that I've copied the code for Alice because now I want Alice to run on my local machine or my local network. So there are a few changes that we will have to make which is we will have to add a name.

So I'm going to call this agent local agent. Then you can also assign a seed to your agent. It's not compulsory. But if you want your agent to have the same address, you can assign a seed phrase which can be a bunch of random characters or some words which are unique to your agent. So I'm going to say my first local agent.

And then I'm just adding a few random numbers to make sure that this is a unique this will assign a unique address to my agent. And then we have to assign a port for the agent so that the agent knows that which port is it is supposed to listen to. So I'm going to assign it to 8003. And then finally we have something known as endpoint. So if you're creating a local agent, you would define an endpoint such as this one uh which is http colon http/local host colon the port number which is 8003 and /submit.

So if you have two agents that are running on your local machine and you want them to communicate, you would have the input enabled. But now I want my local agent to actually talk with the agent that we have which is Bob on agent words. So I'm going to actually replace this with mailbox. So connecting through mailbox is very simple. All that you have to say is mailbox equals true.

And this will allow my agent to send a message from my machine to an agent on a course. So I'm simply going to clear this. And now I'm going to run my agent. So speed local agent. Okay, I did not save my file.

Let me try again. Now we can see that the agent has started with this particular address which ends with uld and it says just about to send a message to Bob and it has given us the agent inspector link. So I will use this agent inspector link to connect to agent worse. So all that you have to do is click on connect and then click on next and that's it. You say okay got it.

And now if you see it will have sent a message to Bob. So if we go to Asian Bob, so the alltime interactions have already gone up to two. And we can also see over here that it has received a message. Hello there Bob B. But it could not send a message back to our agent because at this point we had not connected it to agent first.

So what I'm going to do is I'm going to stop this agent and then I'm going to restart this agent. So that now it's completely connected to mailbox and now it will be able to send a message and also receive back a message from Bob. So if you see this time we've got let's see first if Bob has received a message. Yeah Bob has received a message which says hello there Bob. It's processing the envelope and it is sending back an envelope to LD.

So if we go back and we can see here that it says received a message from hello and the message says hello there friend. So this is how two agents can communicate with each other. If you have one agent on your local machine and if you have another agent on agent words this is how they can communicate. So that was mostly it about the hands-on section. Okay, there is one more thing that I would like to show which is if we go to the agent inspector link this time we can see the messages that were sent and received by our agent.

So you can see that there were two messages that were sent and it only received part one. So this is how you can check the different stats associated with your agent. Now let's go ahead. Now let's say you have uh you have some agents or you want to create agents using lang chain or QAI or MCP. So we have something known as the U agents adapter.

It's a package within new agents which you can use to register these agents on agent course. So if we go to one of these guides here on the innovation lab resources. So this is the guide that will help you register a land graph a course. So all that so this is the standard code for lang chain or a lang graph agent and all that you have to add is just this bit here. So you pass in the name of the agent that we just saw the port associated a description.

So this description is nothing but uh the readme the description that will go in the agents readme it will have the API token. You will need the agent API token. So I'll quickly show you how to get P inverse API token. So if you click on here and if you click on API keys new API key and then you select the right permissions and that's how you can get your API key and then you need to set mailbox equals to true because we are assuming here that your agent would be running on your local machine. Similarly we have an adapter for Q AI which is very similar.

you have to just do the same bits and now there are a lot of people who are using MCPs or building MCP servers. So let's say you want to build a MCP server and you want to host that on a course and you want that to be discoverable through ASI1. Yeah. So just one quick note here. So whenever you have let's say Langraph or QAI agents and you use the U agents adapter those agents will directly have the Asian chat protocol enabled in it.

So this is a great example here which you can refer to and this agent will already have the chat protocol enabled. Now let's say if you want to build using MCPs or you want to create MCP servers on agent words we have an adapter for that as well but at this point of time or you can only create fast MCP servers and with just a few lines of code. So you can have your normal uh MCP server file as server.pi pi py and then an Asian file and all that you need is just this bit here. So you will have to pass in the object of your fast MCP the ASI1 API key.

So you can very easily get the ASI1 API key and you can also decide which model you want to use which is ASI1 mini or extended or fast based on your age. So this is how you can in all get different agents uh to agent wars. Okay. Now comes the important bit. So we expect during the hackathon especially we expect your GitHub repos to have a very good reading.

It should have the project name. It should have some description of the project. It should have well- definfined steps to run your project because that really helps us a lot because if we do not have the right instructions to run your project, we may not be able to run it and evaluate it. You can also define use case examples. If there are any special considerations, for example, if you need some specific API keys, it will be great if you can just point us out to the websites from where we can get the API key.

Next, please ensure if you're building agents and they are hosted on agent words, make sure that you add the agents code in your GitHub repo as well as you define the agents address and the agent name in your project your GitHub readme and you must all u categorize your agents as innovation lab agents. So, let me just go back. So to do that, all you have to do is copy this text in this box. And if I just go to one of my agents that we just created. Sorry.

Yeah. So if I go back to overview edit and if I just keep the rest of I just add it here. I say I want to have the innovation lab tag and click on save. So it will add this view tag here which is innovation lab tag which will help us understand that this is coming from a hathon project. And even when you go to market and if you just search for tag colon innovation lab you can see the different agents that were built during hackathons or built by our team and you can also try and use them.

So for example, this climate guard agent here was built during a recent hackathon and it has the chat with agent button, meaning it has the chat agent chat protocol enabled. And yeah, that's that's mostly it. And now if you are bored listening to me, I have a short video for you guys. So do watch it. I'm going to play it.

I'm not sure if the audio would be audible, but let's try. Search and discovery is changing. Generative AI is unlocking endless possibilities, transforming how we interact with digital services. And consumers expect everything to be on demand and fully connected. But in today's world, everything is isolated, causing frustration and wasted time.

Welcome to Fetch AI, the AI powered platform that's revolutionizing the way we connect and interact with digital services, allowing anyone to connect multiple integrations and deploy AI services at scale anytime and anywhere. Imagine Fetch AI as a vibrant city where each service like booking a taxi or managing warehouse inventory is handled by an AI agent. AI agents handle independent specific tasks like a bike service and a coffee shop, but they can team up to create complex solutions like a coffee delivery service on bikes. This decentralization boosts reliability and flexibility, allowing agents to innovate without waiting for central approval. The possibilities are limitless.

Fetch AI provides the entire tech stack you need to build your own workflows. And getting started is easy. You can be up and running quickly with Fetch Aai's powerful developer tools. Explore Fetch AI agent tech and start building the future then. [Music]