Transcription
In this Cloud Code tutorial, you're going to be learning about agent teams and sub agents. And at the end of this video, you're going to have multiple AI agents building [music] your app at the same time. A front-end developer, a document, and a full agent team all running in parallel inside of Cloud Code working at the same time to get your app built.
This is for anybody using Cloud Code who's tired of doing everything in one long chat that eventually starts falling apart [music] because the more things you put in, the less you're going to get out. You're going to install two global agents, deploy your first sub agent, turn on your agent team so they can coordinate with [music] each other, and then you're going to watch them all work live in VS Code.
Now, this is going to be very important that you grab your SOPs in the description below so you can go ahead and build along with us, right? So, without further ado, let's jump right in. We're going to go straight into installing our first agent. So, if you're following along, and you can also pull up your SOP. So, if you go down to your SOP, your SOP right at the bottom will have all this guide of basically how to do all of this.
So, we're going to go in a little bit of an order. I am going to go through it very quickly because a lot of this once you do it once, you're going to be able to do it again very easily. I'm also going to show you where to get pre-made agents that you have to scan. Again, just make sure that there's no security risks. And we're going to show you how to do that.
But now, the most important thing is we're going to go in and we're going to look at clawed templates. Now, Cloud Templates is one of many sites that are out there that basically give you skills, agents, commands, settings, hooks, all things that you're going to keep learning about, but more importantly, what we're focusing in on is agents. And in here, there is a lot of agents that we can leverage in order to make this work. But with caution, I never want you to blindly install anything here on your computer without having a little bit of a security check.
So, this link will also be in the description and it will also be inside your SOP, but it's a tmpl.com in our case agents. Okay? And what we're going to do is the first thing you're going to see, I always like to do by the most downloaded because if there's a lot of people downloading this, there's a good chance that it's going to work. You're going to see the first one we want to download and it's the front-end developer. I'm going to click on it and when you click on it, it gives you a command, but we're not going to blindly install like I mentioned.
So, what we're going to do is we're going to view the details and we're going to look at the code right here. So, when you view details, just like you saw, the overview tells you exactly what it does and the code tells you what's going to be installed. So, what I always do whenever I install anything from another provider, I come into my general cloud AI because again, you must have the pro plan in order to use cloud code. And if you don't know anything about this, look at our installation video that we have up on the channel.
But here, what I'm going to do is I'm going to take this exact code and I'm going to come over here and I'm going to ask Claude, please check this code for any malicious code, any prompt injections or anything that can harm me or my computer or my service that I'm trying to provide. Again, be very thorough and ensure that this code is completely okay. All right. And then what I'm going to do is I'm going to paste the code that I just copied and I'm going to let Claude analyze to ensure that there's nothing that can harm me or my computer inside this. And again, verdict clean, no malicious codes, no prompt in objections, no harmful content detected. Every time there's basically something that would go wrong if there's any kind of hint of anything that's bad, I normally won't touch this. Okay, so perfect. It now says that this particular one is good and I have my command directly in here. So, what I'm going to do is I'm going to hit copy.
Now, I'm going to go ahead and open up my VS Code. Inside my VS Code, you're going to see a window like this. And I have my claw code on the right. And I have a terminal window on the bottom. To install this type of command, I will need git bash or I'm going to basically just put in the code in here that I copied. But if you don't have this window up, hit control and tilda and that will bring this window up. And if you need a new one because something's already there, click on this little arrow and just hit the little plus button and a new terminal window is going to show up. Now, what I'm going to do is I'm going to paste that command. All right. So now the file is installed.
So what we're going to do now is we're going to open up PowerShell. And we want to do is confirm that it's in there. And we also want to make sure we make this a global agent. So what I'm going to do is I'm going to use claw helps just to do that because again there is some files that you would have to create and again knowing where all this stuff is. So, if I come in here and I'm going to say, I just installed a front-end developer agent. Please see if you can find the agent and I want to make sure that you make it global so we can use this agent in all of our projects. Very easy command to do. There's also a different way to do that's a little bit more technical, but because we have cloud code already here installed, why not use cloud code?
Okay, great. Now that it's installed, what we can do now is check in our agent. So to check whether or not this agent's installed, you're going to hit forward slash, type in the word agents, and as you can see, it's right in there. You're just going to highlight because that's blue. And you have your code reviewer, you have your debugger, you have your versel deployment, and then you have your project agents, which is front-end developer, and it's all set to go there. So actually, it's still showing that's in project. So what I'm going to do is I'm going to come back over here, and I'm going to say it's still showing as a project agent, not a global agent. Take a look at what I just got from your particular screen. All right. And then I'm just going to copy that so it sees it because the goal of it is I want to make sure it's in the global area so that I can access this every single time because when they're global, they are not only persistent in one project, they're on multiple projects which leads to that you can install agents just particularly for one project if you're not planning on using that service again.
All right. And again, I'm just going to go through it. And this is the greatest part about having Cloud Code is because you can make these adjustments directly inside of here. And there we go. It's all installed in there. Super easy to do, right?
So, the next thing that we're going to do is we're going to actually try to before we move forward. We're going to come over here. We're going to go to file. We're going to open folder. And I'll show you where all the agents live. When you go into ULSE, and this is going to be on a PC, very different for Mac, but the instructions that we just did will also work the same for Mac. I'm going to go to my local drive. I'm going to go over to my users folder, click on me, click on Claude, and here is where the agents live. So, I'm going to go ahead and select that folder. So, going forward, when we deploy the next agent, it'll be directly in there, and we won't have to do that extra step.
But the next step is documentation. Now, for whatever reason, believe it or not, documentation is a very important part as you keep on going in and developing apps. What we're going to be doing installing an agent that's going to help us document all of our projects so that it has clear directions on exactly what to do. And honestly, it's probably one of the most important things as you get more and more into this that you want to do, especially if you're having teams help or you're trying to deploy this. There's always something that you're going to need like a document. Whether it's a how-to and how to use the app or anything else that comes along, especially directions for other programmers or developers as again you get bigger and bigger, right?
So, what we're going to do is same thing. We're going to come over here and we are going to go over to our site where we had cloud code templates and we're going to go back to components. Come over here and again I can make my own agent too and I'm going to show you how to do that in a minute. But what we're going to do is we're going to go to agents and we're going to look for anything in here that really does it. But what I want to do is look mobile developer. That's going to come pretty cool. But there is one in here pretty much that I really like the way it was written. You have a technical writer, you have a database, but you have a documentation expert. This is going to be the one that I want.
And again, just like before, I click on view details. I go to code. And then here, I copy the code, go into my cloud code AI, and I'm going to just validate that the code I'm about to put into my system does not have any malicious code. And it's a very simple one, but it's very powerful. And the reason why we need to document is again, this is going to become incredibly in handy later on, especially if you're building an app that you want to have directions to. How-to guides. This will be amazing for anybody to have in any of their systems. Right? So, it's looking clean, safe to use. I got the green light.
And again, this is a new window, per se. I'm going to hit control tilda. This is going to come up. I need to like have this installed. So, all I'm going to do is Ctrl +V. This is Git Bash. If you don't know what Git Bash is, watch the first install video. And I'm going to come in here. Boom. And that's it. But look, this time around, we made it super easy. It should have put it in there. Don't know why it's not in there, but we're going to validate. There's agents. And then it should have been an agent. So, it's one file below. But again, to make it super easy, I'm just going to move it myself. I'm going to come in here into my documents folder. That was my mistake. I should have done it. But I want you guys to experience this because if you run into the same issues, you always want to make these things global and you don't want to have any issues, right? So, it is in agents and it is in clawed. So, it double created a file. All I'm going to do is Ctrl + C or actually you can use the cut command right in here by right-clicking it. Hit cut. Go back up one level. Paste it. There it is right here. And I'm just going to delete this other folder because it didn't need to make it. So going forward, instead of being in that one, we should have been inside agents, but it automatically created cloud code. Not a big deal. We just moved it on over, but now it's here. And if you ever want to check it now, what I'll do is I'll write in Claude. Come in here. Let Claude comes in. Yes, I trust this folder. And I'm just going to type in agents, manage agent configurations. And we should Oh, we should have it directly in here. Documentation expert. And it says inherit. That means it's going to inherit whatever the main model is that we're going to be using. So that's why that's there.
Okay. So now you had the ability of creating all of this. What if we can create our own agent? What if I wanted to create a copywriting agent? Or what if I wanted to create an agent that does something else? Or what if I wanted to modify the code that's there and make it even better based on my own personal experience. Right?
So, let's get into creating a new agent. So, what we're going to do is I'm going to do this completely from scratch. I'm going to hit escape. I'm also going to go ahead and hit and type in the word exit. So, we can come out of this al together. And what I'm going to do is I'm going to come over here and I'm going to do a new window, a new terminal window. In here, I'm going to type in claude. The claude window comes up. And once it comes up, I'm going to type in agent. And when I hit agent, it's going to have this particularly like light blue kind of highlighted. And I'm going to hit create a new agent. Okay. Project means for this particular project. Personal means that it's going to make it global. I'm going to make it global. And we're going to generate you with claude or we're going to do manual configuration. I'm going to do manual configuration so I can show you how to do this. So again, this is completely unlimited.
Then I'm also going to show you a little bit of a hack that you want to try if you see a great agent, but maybe it's in a not so trusty like cloud template so far has proven to be really trustworthy and it's just a really good source to get there. But maybe I want to do something a little bit different. So I'm going to go back to components, right? Let's go to agents and I'm going to type in copywriting. Okay. And look, there's a copywriting skill right in here. But I'm going to view details. I'm going to look at the code. Copy the code here. I'm going to go to Claude Code. I'm going to open up a new chat window.
Now, if you've been using any of these Claude code, chatgp, and you've been writing marketing content for yourself or newsletter or anything else, it might know your writing style. So, what I'll tend to do is I'm going to say, I'm creating an agent that I want to ensure has my copywriting style based on everything that you know about me and all the things that you helped me create in the past. So, can you take this current claude code code for an agent and modify it towards my personal copywriting skill and let's name it Nunos's copywriter so that way there's no confusion if there's another copywriting agent there. And remember, the main goal of this is to create an agent. Also, ensure that you give me things in an order that I can copy and paste to the agent creator. So, therefore, it doesn't give me any problems when I'm creating my agents. All right.
So, what I'm going to do is I wrote that little prompt. Well, I didn't write it. I spoke it. And I'm going to hit enter. And what it's going to do, it's basically going to give me the code that I'm going to need. Now, I am going to pause this video because this might take a little bit. All right.
Now, what's really cool about this is that it gave me in the exact kind of way that we're going to need it. And I'll show you what I mean. So, we're going to go in VS Code, right? It's going to enter a unique identifier for your agent, like whatever the name is going to be. And again, if you come back over to our area here that we just had. All right, it's going to be Nuno's copyriter. So, we're just going to copy that. We're going to go to our terminal window. Let me make this a little bit smaller. So, then go in. Let me just squeeze this a little bit. There we go. Let's get rid of this window because we need our chat window. And the unique identifier is going to be Nuno's copyriter. Oh, no spaces. So, copywriter. Again, we're in the terminal. So, we can't use the mouse. We're just going to put Nun knows copywriter. Okay, we're going to hit enter. Enter the system prompt be used for the agent. That is going to be this right here. It's going to be the entire prompt. You see that? And then let's go back over to our terminal again. We're going to paste in the prompt. Hit enter. And then when should claude code be using this. So usually it gives that's in the description. But again, if you get like a little confused, all you got to do is cloud code just asked me when should this agent be used? What should I write? And again, just make it super simple. You have AI to help you create AI. I mean, literally, we're living in the matrix. It's so freaking cool. All right, there we go. C. Come back over here. Go to our terminal window again. Paste it. And then it's going to ask what tools. Again, you could have all these in here or you can have it where it's read only and everything else. In this particular one, I'm going to go ahead and hit all tools because it's definitely going to be just for me. Now, for complex reasoning writing, I feel that sonnet is pretty good. If you want to keep it cheaper, you go to ha coup or you inherit from the parent, whatever it is. We're going to keep it at sonnet because again, it's copyrightiting. It's going to need its most creativity and most advanced model in my opinion. And I'm going to hit enter and we're going to give it a color. We'll give this one orange. And we're going to enable clawed agent memory. Before this was something that was not available. Now, in the recent updates, it's pretty good. So, we're going to go ahead and enable memory. We're going to go ahead and hit enter and save. And congratulations. We just created a copywriter agent all by yourself.
We installed our document. We did the agent file. We talked about like, you know, the different formatting and all that. But now, we're going to talk about what a sub agent is and what does it do. Now, if you remember, we created a document reviewer, right? And what we can do in any of this point in time is we now can use our agent to document our projects. Now, I have a couple projects open. I'm going to use the document as a single sub agent to go get stuff done for me because it's already trained and it's an expert in that particular area.
Now, why this is so important is later on when you're getting into deployment and you're getting into database, you can create a database agent. You can create a deployment agent that works specifically with very specific tools and helping you deploy. Now, I'm going to come in here and what I'm going to do is I'm going to go to file. I'm gonna open a folder because what I want to do is open my one of my projects that I have inside of Claw Code. So, I'll go to my projects folder and I have a book creator that I'm doing. I'm going to double click on that book creator. I'm just going to select the folder.
What I'm going to do is I can do it one of two ways. One, I can come over here and I can use my Claude Code Assistant or I can go directly into the terminal. Kind of up to you. All right. Now, what I'll do is I opened up that project folder. I'm going to go down here to the terminal. I'm going to create a new terminal window and and you don't have to do that. I'm just doing it so then it's easier. I'm just going to type in the word claw to start claw code directly in here. And what I'm going to do is I'm going to invoke my agent. You can do it multiple ways. One, you can just speak it inside of your area in here and you can make it work. Or two, you can actually physically call the agent. So if you come in here, I can go ahead and start writing documentation expert. Hit enter. And I'm going to say please create the documents for this project. I'll let it do its own thing because it's now it's going to ask me a bunch of questions about the project, what I want, and everything else. It's going to automatically go ahead and deploy this thing, which is going to make it super cool, right? And then what it's going to do is it's going to create different things that I might need and so forth and so on.
If I was building a front-end development team, like if I wanted to create a website, we would use a front-end developer. Just get you an idea of how to do this. Now, imagine you doing this every single time you need something. you're going to have an expert on hand that you can call or that will automatically turn on because the system knows that it has that documentation expert because what it's waiting for is that particular description that we added of when to use this. If I wanted to use the copywriting one and I started writing copy for let's say a front-end page, it would automatically call that agent for me just speaking into it. But this way, this is something specific that I want. I'm calling that agent in order to do so. And then again, we'll go back in and we'll do it a little bit differently with agent teams when we turn that feature on so you can see it. But now we're physically calling it. So you can call an agent. You can also just start the project and the agent will automatically activate if it feels needs the agent to do that task. And again, when you use agent teams, it will deploy these agents automatically based on specifications of the project when you're writing your claw MD file or your PRD for the particular project.
All right. And here you go. Look, it's got all these crazy things that it just added. It added MD files. Here it is. So, you can see it written. This is like everything that it has in there. So, it created all these already. And now you have it. Look, here's your readme file, what the app is about, what it's doing, how it works, and everything else. And if I even wanted to is could you also create a how-to guide for me for my users? And I can create a little how-to guide for the users that are going to have it that I can use later on at any point in time.
All right. So, we created a sub agent. it went in. And what's great about this is if you're working on multiple projects, like for instance, I could be working on another project, create another window for that same project, work with a different sub agent, and that agent can work in parallel. You can have multiple agents running at the same time for any individual project. And what's important about this is is that each agent that you're going to invoke or start will have its own context window, which means that they will be operating independently and reporting back to your main orchestrator. So for instance, what do I mean by that? If I wanted to create a brand new website or a project that I wanted to document at the same time, I could say to the AI, please create the following project. Use the following agents, this agent, this agent, and this agent. And create the following project using this PRD file and whatever it is that you need or the prompt that you're going to give it. And what it will do is it will spawn three different agents to go ahead and create your particular project. Pretty darn cool.
Right now, the reason why that's so important and why you have to get used to building it this way is because this reduces the amount of money you spend on tokens because what will happen is each individual agent is using its own memory. So, it doesn't always have to recap everything from the previous, you know, prompt that you gave it. And the reason that's important is because then your cost per tokens is going to be lower and you're going to have better output because they have a very confined memory per each agent that's being used. So agents is right now the future of 2026 and beyond because if you get a hold of agents and you start working with them the right way, you're able to deploy a mass amount of agents doing a lot of different jobs.
Now what's really cool about what Claude has done is it released something called agent teams. Now before when you're deploying sub agents you have the main prompt which you're working out of plus each agent is working independently. The thing about that is that each agent when it completes the project has to communicate back that it's done and all these agents sometimes don't see what the other agents are doing. Not bad if one is a document and the other one is a builder. But it will get kind of weird if like your front-end developer makes something that the back-end developer can't see and it causes problems later on that your main kind of like manager needs to figure out, right?
So what agent teams is is instead of the agent teams working independently, they all start to communicate with each other. It's an experimental feature that we have to turn on. But be aware that token costs can range two to seven times higher if the agents start talking to each other and the developing the product. Now you're going to say, well, why do I want that? If I rather do single agent teams in order to get everything and keep my cost low, not necessarily. The benefit about having a deployed agent team is that it will automatically create the agents that it will need. the agents will communicate with each other and believe it or not your app will build incredibly faster. So the advantage of agent teams versus single sub agents working in parallel and independently is that the agent teams all talk to each other and it produces a better and faster result at least as of this video.
Now later on you will learn even more cost-saving tips on deploying agent teams but where the agent teams are using different models to keep your cost fairly low. But that's a super advanced kind of area because you have to program that and there's a flow that you have to kind of develop. But to start off with if I needed an app in a rush or in a hurry and I wanted a full complete app I would go in and I would use agent teams in order to do that even though my cost would be a little higher. Now, if you're on the Pro Max plan with Claude, you should be okay because your subscription will hold quite a bit. I pay the $200 a month version. And I got to be honest, I usually stay within those guidelines. I don't blow out through my credits, but every now and again, I do work on pretty significant projects and my cloud code expenses do go a little high, right? But that's something to think about.
So, hopefully you understand. But what we're going to do now is we're going to turn on agent teams. So, there's a couple ways you can do it. The easiest way honestly is to copy and paste this command right here that you're going to have in an SOP because you would want it to just automatically go in there. Or you can physically change and modify the JSON file in your cloud settings folder. I personally don't like messing around with the JSON file, but I have cloud code with AI, so why would I have to do that?
So, the first thing that we're going to do is we're going to copy the cloud code. We're going to want to make sure that our cloud code is updated to at least 2.1.32. So I'm going to come over here and I'm going to just copy this command real quick which is just validating that if there's an update updated and just to check on my version. So I'm going to come over here to my terminal. I'm going to put clear and it clears the terminal al together. I'm going to do the new bash window. I'm going to come over here. This is technically creating a new like instance. I'm going to hit control +v and I'm just going to ensure that I have it. See, it says 2.1.63. Yours is my much higher number when this video gets released, but just keep that in mind. You need to have that certain version.
All right. Now, the next thing I'm going to do is type in Claude to get Claude going. And I'm going to go back into my presentation here. And I'm just going to go ahead and copy this right here, which is add the agent teams experimental flag to my project settings. Add cloud experimental agents set to one environmental settings in all of this. But what I'm going to do is I'm just going to come over here. I would just basically paste. But instead of doing that because I've already turned this feature on, what I'm going to do is can you please ensure that I have agent teams set. It's an experimental feature in the claude settings. Can you please ensure that that's on and that's working globally for me. I'm just going to do it because again, I've already installed it. You would install using that code or you're going to go in and you're going to turn this in your JSON file in your cloud settings.
Now again, it's not very terrifying. If you need to look at the folder, you come in here. You go to again your C drive or in your Apple computer, look for your cla settings folder. You're going to come over here. You're going to go over to users. You're going to go in here to the file. And then in here, you're going to notice there is a cloud file, which is your JSON source file. And you can open it up with Visual Studio. I mess around in here. Again, I'm not going to do that. I'm going to use the very smart system that I have in here, which makes it really nice. and agent teams is already enabled in your global settings on line three. You can see it here. The environment set to one. I'm ready to good to go. If not, it will ask you a whole bunch of like can I change this file settings and you can go ahead and do that. But that's how easy it is, guys.
Now that that system is on, we can invoke it by doing it in a certain way. Right now, I could say create an agent team to review my project. All right. And we'll do this. And what it's going to do is it's going to create a set of agents that's going to basically review this particular object. And what it's going to do, it's going to basically go out there and it's going to check this project out and it's going to create the agents that it feels it needs. So it's going to ask me, do I want to proceed, of course, and it's going to ask me some just like basic permissions in order for things to get done. All right. And it's going to go in here. What's really cool about it is is that you'll be able to go through it and it's telling you everything that it's going to do. All right. And if it feels that there's already a pre-created agent that it needs to use or invoke, it will go ahead and do that right now. Again, it's a bunch of commands. I'm not going to go through all of them just because I've been through it and we're trying to make this video short. And we're going through it and it's going into plan mode. It's going to basically go in and it's going to check everything out and it's running in the background.
Now, all this is pretty cool. Later on, we're actually going to use another project and we're going to see we're going to enable an extension. So, hopefully we could see our agents work, which is going to be pretty cool. But we're going to give this a second to go through the thing. And now it's going to say, let me create a review task and spawn team members to work in parallel. Review fun front end code, review backend code, quality, and security. These are all the different project teams or the team members that it's going to invoke in order to do that. And then they're going to run all three reviews in parallel. And it's going to do it all on its own. Pretty cool. We literally just unleashed an army to finish and to ensure that this project is good. And again, look, it's using the front-end reviewer. It created that particular person for this particular project. And this thing is going to go on. Now, again, remember, you're probably going to have a little bit more higher token cost because you are invoking it to do its own agents and so forth and so on. But look, here it is. It's launched. It's reviewing its thing. We're going to let this thing cook for a little bit and it's automatically running, which I think is pretty darn cool.
So, here you can actually see that. Look, it's doing all these things and it says press the down arrow to extend or control T to show tasks. So, we're going to hit CtrlT and it's showing the tasks so you could see everything that's going on and you can expand and you could see what each other one is doing and you can message them individually. How cool is that, right? So, this is how neat this thing is. You could see this is the main which is the main area right here. You could see what this particular messenger is doing. And you'll notice that it has different colors based on who's doing what at what given point in time, which I think is pretty cool. And it's got the team lead that's running the whole entire thing. All right.
Now, it said that it's got a full review. There were five criticals, 21 warnings, and 16 suggestions across front end, backend, and architecture. Would you like me to start fixing the issues? I recommend tackling the five critical issues first. And I'm going to say absolutely. Just come in here. Yes. All right. and it would do that. And again, it would go ahead and just build all of that. Right?
So, pretty cool. Right now, let's go back to our presentation. The next thing that we're going to do is we've launched our team. It did all that. Again, you can create as many agents as you can. Now, the really cool thing about this is is that you're going to go in and when you build these, the more agents you build, the better and more efficient they'll become because again, based on the skill sets that you're giving them, the MCPs and the connections. And if you don't know what that means, go watch the previous videos that I've done on this, which I go deep into detail into those because again, this is kind of a series. This is like part three and there is going to be a part four where we're going to learn about commands and hooks because those are very important. And we're going to just keep going down this pike of understanding how to do clot code and how to work with it because I honestly do believe that this is the future.
Now, what we're going to do is install a visual way for us to see our agents work. So, we're going to use agent snitch. There's a couple of other ones that you might want to try. We find that this has got a fivestar review and it works with like color codeex and a bunch of other ones. So we're going to go ahead and install. All right. And we're going to trust the publisher. And what it does is it basically creates like this visual component of making it work. So what we're going to do is we're going to open install open claw code using your agents as normal and open snitch view which is now on the sidebar icon which is this thing right here. So we can basically see what's going on. Right.
So, what we're going to do in the middle is we're going to open up our Well, first let's go into another folder. So, I'm going to open up a folder, Powerball. I'm going to hit select folder. This is a fun one that I just created for myself. What I'm going to do is I'm going to open up here. And just like before, I'm going to go to Claude. I'm going to open up Claude itself in the terminal. Okay. Yes, I trust this folder. And I have agent stitch involved. And I'm going to say create a team of agents to review this project. All right.
So, hopefully this thing will invoke. And look at that. You're now seeing Claude going in there and it's doing the different things. It's trying different commands, but it's having a right [laughter] an issue whatsoever in there, but it's working through and you're starting to see all the different things it's starting to do, which is pretty cool. It's creating the team. So, you can visually see what's going on there. A pretty cool one, there's ones that you could do the pixel ones that is out there, but you have to configure it a little bit. But how cool is this? Now, you can see all your agents work and you can see the different projects it's creating to make it make a lot more sense for you guys.
So let's recap everything that we did. We created basically two agents from scratch. We used claude code templates in order to download those agents. Once we did that, we also created our own agents so we can see how it can do it from absolute scratch. We then learned about sub agents and how to run these sub agents to do individual tasks. We then took that and we turned on the agent team experimental feature. Once we did that, we deployed a group of agents. We also learned that agent teams and single agents have different costs, right? The agent teams are going to cost a little bit more because they're all talking to each other. And then finally, the last thing that we did is came in here and we started looking at it visually see what's happening with inside our own cloud code and seeing what different things are happening as different tasks are created and our team are being deployed.
So, make sure you stay tuned as we go over commands and hooks on our next video, which again is starting to go in through our whole series of how to use claw code. And I hope you're enjoying this series. If anything, make sure to subscribe and like this channel. And we'll see you in the next one.