How to Build AI Agents that Actually Work… (NO CODE)

I'm going to show you exactly how to build AI agents that actually work. AI can be very confusing and there are so many different concepts to understand. And after spending literally hundreds of hours building AI agents, I'm going to show you exactly no fluff. How to build AI agents easily with no code across three levels that you need to master so you can build digital employees and bring any idea to life whilst avoiding the very costly mistakes. [music] Even if you're a complete beginner. And if you're new, I'm Jack. I built and sold my last tech startup with like a galion customers. Now I'm building my own AI startups and I share my channel with stuff that actually work. So if you haven't already, grab that beautiful coffee. And let's dive straight in. So we need to build this across three levels. Access, knowledge, and proactivity. So level one is going to be access. It's an agent that essentially lives where work happens. Now we're going to be building this agent inside something called hyper agent. You reached out and said, "Jack, we would love to do a video with you." And I've used Air Table for a while. So, I was very stoked to check out the platform. So, thank you for you guys for sponsoring this video. And they've also given you $500 of free credits for the first I think it's 500 people. So, I'll put a link down below for that. Um, but whether you're building this in any agent system, and I'm going to show you how Hyper Agent's different and where it's sort of unique. Um, but whatever you're building in, these principles will apply to literally any agent system. So, it's very worthwhile knowing and important to understand. Now, the idea with access is we essentially, let's say we have a team, okay, we got a whole team here. I don't know who all these guys are with their beautiful mustaches, but they're looking very fantastic if you ask me. Essentially, let's go ahead and build something, okay, that actually solves a business problem or a personal problem that I can access, maybe my team can access, or anybody in your organization can physically do. So, I'm going to sh over now to hyper agent to go ahead and do this. And what I want it to do realistically is to let's say I want it to build for me a brand guidelines, okay? Brand guidelines for a website. So I might say, "Hey there, I want you to go over to glido.com and I would like you to do two things. I want you to extract the brand guidelines. So tell me like what are the features of the brand and then just build for me a two-page HTML presentation explaining its core features." Okay. So I'm going to come over here. We got different modes. Obviously you got plan if you want to think through a bit more or it can execute. I'm going to go ahead and want it to create a plan. So I'm just going to send this one off. And then what we're going to do is actually turn this into an agent that I can access in tag wherever I am. Whether that's in Telegram or Slack or let's say I have my team in Slack. I have five people and I build an agent that can do something really cool. Well, we can actually get all get access to this agent together. So, for example, my thumbnail designer could talk to it. My video editor could talk to it and we can build up these individual expertise. And whilst hyper agent is working in the background, access level is all about thinking about where does work happen for you. You shouldn't always need to add in 17 different platforms. Think about where am I actually physically doing work. Am I doing it in Slack? Am I doing it in Telegram? And when we build our agents, we want to get them to where work is happening. It's asking us some follow questions. Let's do a two slide deck. Uh 16x9 navigable is fantastic. Come down and let's just do brand identity. That's cool. And then send off the response. When this is complete, it just asks us to approve the plan. We can see the thread context on the right hand side. Go down approve it. And then we'll send this one off. And as you can see how it's reasoning in the background. It's actually gone ahead and found the glider website. It's reasoning through it. You can send it off. It will action it off. But the magic happens realistically once you've turned it into an agent and you can call it and ask it anything. And just like that, guys, it's done. Now, one interesting thing I want to point out about the way that this agent's reasoning is you can kind of see all the stuff that it's doing as it's going along, which is really cool. So, we've got the HTML deck here. Awesome guys. And it's done. Now, one thing I want to call out is that the way they've set this up is effectively um it shows its reasoning and you have these different threads. You can check it out and tiles and find out the thing that you want. And so this is what it built the glider brand identity. So let's go ahead and full screen this bad boy. And guys look at this. I actually think this is freaking gorgeous design. It's crushes. And bar this is one shot. I give it I just told it one thing. I come over and that looks pretty fantastic. So let me exit this just so you can get a better view. If I come down to the carousel view, you can come over and I click on this. I'm just going to full screen it right here. You can see what I'm talking about. Refresh. Good animations. Come down. We're rocking. We're rolling. We come over. Really simple task. But we don't just want to give it oneoff task. What we want to do is turn it into a reusable agent. Beautiful. And so to turn it into an agent, I'm just going to say, "Hey, I would like to turn this into an agent that I can use in my Slack channel, make it a persistent agent that I can call." Couldn't come down and send that one off. And they will literally guide us through that system. We can also set up a live agent. I'll show you how that works later where basically it can proactively reach out to you and do some epic stuff. And so guys, this is what I like about the fact I can tell it to do something and it does. It's basically gone ahead and hey, this is a brand agent. I'm going to come down and click on save. And if we're unsure, it can actually guide us through on how to set things up. I know it's become an agent because the color in the background actually changes. And so now we can actually see the agent once you click on agents. Now this is one of the cool things that Hyperent does. So the idea is like if again if you have a team of like 20 30 people, everybody can call the same agent that you set up. So we have the brand lens that describes what this does. I want to come down here to invocations real quick. So say for example, I want to connect it. I'm going to go ahead and choose Telegram. So go ahead and connect to Telegram real quick. So basically we can create a bot on Telegram that we can all chat to and ask questions to. So we come over to the bot father. We do for/ newbot like so. So then we want to give it a name. I'm going to come down here and give it this one. Let's do a quick underscorebot very quickly. So you're going to get a token guys. I want to go ahead and literally copy this token like so. And then come back over here and drop it in here. And as you can see it's now appeared. It's ready to rock and roll and it's live. And just click on connect. And then we're going to pair this with Telegram. But let's go back over now to Telegram. And then if I click on this I can actually drop in a message. So click on this guy here. Beautiful. Now let's go back over to Hyper Agent like so and click on open telegram to pair. Then come down and select open telegram. And just like that you have paired with a hyper agent. You can now chat with your agent here. So what's cool about this is you can literally get anybody to message this specific bot now in your team. I said dude you need this thing just drop this guy a message and everybody can access it. So I might say hey could you do me a favor? Go ahead and create for me a two-pager uh HTML presentation on stripe.com. Okay beautiful. So, I could say maybe I'm working with Stripe. Maybe I'm working with the UFC. Hopefully, you saw the card this weekend. Crazy crazy night fight. It's really enjoyed it. Um, but again, we can ask it these questions. It'll now go ahead and do this. And this could be anybody messaging this. Beautiful. So, now our agent has finished. It's gone ahead and it's done all the work and it's got this right brand lens. We're going to see this together right now for the first time to see how it's done. I'm going to click on this bad boy. It's opened it up. Made with Hyper Agent and wow. Do you know genuinely, I'm I'm impressed with that for a one shot. Obviously, it's the model for 4.8, eight. But just to have created that on something and then have everyone in the team to do it, we could perfect it and then just have our team do all the cool stuff with it. Really cool. Yeah. Really fresh. And again, this is tasty. Tasty being the operative word. So now we basically got our agents where our work is. That could be Telegram, could be Slack. But then we need to take this to a new level. So we have an agent now that everybody can message. And we've accomplished that. The next thing that we need to do is give it critical knowledge. Now, if it's something that you're using, it needs to know about you, who you are, your strategy, access to your connections, your emails, your calendars, or if you're a business, a company, an enterprise, it needs to know about your policies and how all of that connects together. And by this sounds like I'm speaking Spanish, I'll put a link down below for the full Claude called Code Masterass that explains all about uh setting up AI. It takes you through foundation setups, building websites, power features, memory systems, all the way through down to monetization and a gargantillion other things as well as our full claw code operating system if this sounds a bit like Spanish now and you want to get way far ahead. Now, here's the thing with knowledge, right? If we come back over now to hyper agent in this case, and again, these same principles apply to any agent system that you're going to be using. What we can do is we can create a specific agent that we chat to about certain things. So, for example, let's say that we want to have a HR agent or you're building this for a company. What we could do if we wanted to is create a brand new HR agent or we could even add what we call universal memories. For example, if you come down here on the left hand side to memories, what we can do here is add in certain memories. For example, if I click on add my first memory, I might say, hey, Jack loves MMA specifically working a lot on striking, spinning back kicks right now and setups. Whatever the thing is, we can set the importance of that. I might say, hey, you know, always include this is like a level five important piece of information. It's a user fact, whatever the the hell the thing is. When to use optional, I could say something like um whenever asked a question about Jack and I could just say something like preferences. So this is small snack size stuff. Uh in Hermes, it's basically just like you tell it, hey, remember this and it will commit to his memory. This has a full systemized way of doing it. Scope is global. So we can also scope this to agent. So I say, hey, this is something I I want all agents to know. I can come down, I can add the memory. Now, if I go through to Telegram, I can come and say, "Hey, um, tell me an interesting fact about Jack." Okay? And this is just show you how it works. And when the memory system is really, really important and any agents that you're using and just like that, a genuinely interesting fact about Jack is the spinning back kick. Very, very, very cool. So, what we're going to do now is spin up basically an agent that has lots of information. So, what I could do, for example, if I click on a, let's say, a news thread, I could say something like, "Hey there, what I'd like to do is create a company resource agent that people can ask for policies or anything they want to know about time off or whatever it is. Um, what I would love you to do actually just as an instance, we're going to build one for glider.com, a speechtoext app, and I just want you to consume all the information and facts from their web page and add that to your knowledge base. And I'm also going to upload uh an article. I want you to consolidate that too. And this needs to be accessible via Telegram. Also, I've also got a random document talking about 300 days off my HR policies. So, I'm just going to upload that right here. By the way, here are the HR policies. Add this in there also. And when this is done, I want to send this into an agent that basically we can ask to any questions on Telegram. And then send this off. One thing to call out with these agent guys are these are in the cloud. So, in other words, like they will always be accessible and on 24/7. That's part of like the the hyper agent thing, which to me is a is a pretty cool thing that they're always available to do that unless you're running one on your local. It's cool that that they've got these things that always running in the background. Beautiful. So now it has completed and it say, "Hey Jack, do you want me to turn this into a glider resource assistant?" We sure do. We're going to click on save agent. And in exactly the same way as before. Now we can view all of our agents beautifully here on the left hand side. I'm a stickler for renaming these. So you can change them to whatever you want to. But again, we can set up any kind of mode we want to. This is what it has access to. We can connect it to Slack, the Telegram, email aliases, so many different things that we want to do. And every time it has a conversation, you see the thread down below, just like it did with the brand lens, you can see all different conversations will all appear as threads at the bottom, which is super handy. So you can see them. The other cool thing as well, I quite like on transparency because cost management is a really important thing with these as you scale up. Obviously, they've hooked it with $500 credits, which is really cool that they've done that. I we appreciate that. Is that you want to make sure that your costs are under control and you're using the right model for it. You know, more design related stuff, you wouldn't be using GLM 5.2 and your Opus models. Um, but for more communication stuff, I quite like leaning on the DeepSc V4s, the um the GLM 5.2s, the Chat GPT models to kind of keep your costs down. So, that's all really cool. And again, now we have this resource assistant. We can have conversations with it. It has access to tools. We can set those up and do a lot of different things. Now, under here, we have autonomy. Uh, again, you can go through all these different settings and have conversations. This essentially, you know, is it auto? Agent runs everything end to end without stopping or does it ask us first? Really great to have that level of control. We've got ideas on identity. We have things here on the activity that it's going to take the model selection. So, I can literally come down and configure model based on what I want to do. Let's go for GLM 5.2 fast on this one. Again, really cool that we can do that. Then we have our tools and integrations, which is fantastic. And it already has lots of different skills in here. Image gen, video gen, which is cool. And now any invocations. Again, I can connect this to Telegram if I want to. And the idea being you can just have a series of bots. But just to show you how this works, if I come back over to the threads under company knowledge, I can come down and say, "Hey there, just refresh my memory. What is the um holiday the annual leave policy?" It's so funny. I've not worked I left my corporate job to do my last tech startup quite a few years ago, maybe like I don't know what's a specific year, but a few years ago and I I've since forgotten about the annual leave. We had 25 days annual leave and you had to like book it and it was um wild. didn't wasn't quite the 300 days of paid annual leave that you get a glide apparently which is really cool. Interesting fact if you give someone unlimited annual leave they actually take less than 25 days on average which I think is really interesting. But here at this fictitious glider company we got here you get 300 days. So that's exactly how it works and we've connected knowledge to it and a knowledge thing is really important because it isn't just about the knowledge. We want to bring in Gmail and calendar. You'll see in all my Hermes agent videos I talk about the very kind of intricate memory steps that we have and how that's so important. We can upload documents just like that. We can connect loads of different connections here to um hyper agent. So we can actually give different skills and tools to each agent here. Again, this works in every every agent works differently in hyper agent. On the right hand side, if you come down here to tools, we've got integrations, all this sort of stuff. You click on integrations, you can see you can connect it to ampify to Apollo, firecrawl, all the tools that we talk about regularly on the channel. You can just literally connect to it within the video, which is really cool. And then that agent will have those specific integrations. Then once you are connected, you can essentially call that whenever you would like to. And to show it looks like, we're going to go ahead here and connect to Zappia. Now Zapia is like the I would call it the WD40, the Swiss Army knife for the integrations. If you can't connect to it, you connect to Zappia and Zapia will connect to everything. It's really overpowered. I use it literally every day in my business. It's really cool. In fact, one of my first tech startups we built using Zapia in the MVP stage. It was so cool. So literally, you're going to come down. It's going to find your account. You're going to go ahead and click allow. And she is now going ahead and connecting to Zapia. is now fully approved, which is awesome. So, if I were to then go ahead and create a third agent, we're going to use Zapia now to actually use one of its skills to do something really cool and interesting. As you can see now in here under integrations, you can see it now has Zapia, which is fantastic. So, I can go back over to a thread and have that conversation. And literally once you're configured with Zapia, and if you haven't, you can literally head over to this website, which is mcp.zapia.com. You can connect all of your accounts to Zapia. And then here's the really cool thing. So, you can see I've got access to all of these different tools and systems here. You can see I've got all of them which is fantastic. School, Slack, Canva, Paperform, Gumroad, a million things. So I could say something like, "Awesome. I want you to go ahead and access Gumroad and just give me one interesting piece of information from my email system in Gumroad." Go ahead and send that one off and then we can actually get this data. Now, whatever the thing is, we can connect it. We can use Zappia. We can use the integrated connections within Hyper agent or custom MCPS if we want to and get that specific information, which means that all the different agents we create are important. Wherever you're building your agent in any system, you want to make sure that you give it the right tools. What's really really important though is that you give it under the principles of least required access. So never give any specific agent access to anything else that it needs. For example, don't give it the ability to create emails. And also in each of the agents that you do create, whether that's in Hermes, open call, hyper agent, make sure that you kind of spec it to that specific agent and you don't give the keys to the kingdom to absolutely everybody. Beautiful. And now we've got the latest Gum Raid soul. It's not really a sell cuz it's at $0. It's completely free anti-gravity master class and it's even given me the bio. So you imagine you have like a system of knowledge. You can just ask it these questions and all I did is just connected to Zapia and I've already brought over all of my existing connections. And so now we've really thought hard about the knowledge. The third level that we now need to look at once we've gone past that, we've got past a bit of it remembering all that sort of stuff is what I call proactivity. Because us sending a message is one thing, but it's ability to proactively come back and say, "Hey, Jack, here's the thing that I'm giving you without you even asking for it." Proactivity is so important. We call it high agency. Just doing it and making it happen. An idea is we can do this in a couple ways with agents generally speaking. We could do it on a timer. So, an agent, for example, and you've seen on the previous Hermes Masterclass videos I've done that you can get it to set daily briefs and daily reminders, or you could do it when something happens. Now, cool thing with Hyper Agent is they've got loads of ways to make this happen, which is very, very cool. We could do a weekly digest for, let's say, in this example, right? We've got a Telegram bot that we're all talking to. What we could do is every Friday get it to release the sales for the week. So, go over to my Air Table, find out what the sales were, and then just let everybody know what that is on repeat. So, you just get a literal notification from this thing, say, "Hey, dude, here are all the sales that you got." That's how cool this is. And so, for example, I give it this prompt here, which is I would like it to go ahead and create for me a one-page 16x9 onboarding deck welcoming new GLD hires. I want to go ahead and ask the GLD resource agent for the team annually policy and any key company facts. Now, what's incredible here, guys, is we're actually doing agentto agent communication. So, this agent that's doing all this magic here, as you can see on the right hand side, can actually consult with another agent and talk to it by itself whilst also completing all of our actions. And you can even see its actions. It's going through the browser section and it's gotten all that information. And as you can see, guys, after it's reasoned through, which it's done a great job of doing, it's actually pulled together this full thing in our design branding. And crucially, it has crushed this idea that you get 300 days of paid leave every year. So the idea now with these agents in hyper agent is then not only can we use them one by one, but we can also stack them on top of each other and have them reference and speak to each other, but we can take this to a new level. So, for example, let's ask it to turn it into an agent. And what I'm going to do here is make it proactive and actually have it update this on a regular basis. And just like that, we now have the Glideo onboarding studio. It's going to come down and save this agent. Then once I select this agent on the left hand side on this page, if I come down to invocations, what I can do is actually set up live mode by clicking on live mode here or I can click up there. And what's really cool is we can set this to run. Let's say we want this to run every 7 days. For instance, you could do daily morning breathes for anything you want to. is completely up to you and I might give it a prompt like hey there I'd like to go ahead and update the onboarding deck with any changes to the company policy and what this means is anytime it changes we will then update the deck and I can pick the heartbeat model and as I do this say for example I wanted fable 5 the cost actually changes at the bottom so that's $445 but if I come down to for example open 4.8 The cost then goes down. Then we click enable and effectively it is ready to rock and roll. And we can even pick where this happens whether it's in a chat or directly on Telegram or Slack. And so the key things here really come in alignment. What are the big things you've got to do in your business in your personal life and where can you add value? And you basically build up an agent or a team of agents to support you in delivering against that goal. Whether that's a weekly digest, sales figures, anything you're doing manually, we can get a team now to do that and share memories across everybody else. So just to contextualize this into tools that you already know, I thought that'd be really helpful just to cover. So when you're thinking of core code, it runs locally in terminal. It's really good for code and building you're at your computer. Hermes agent itself is a super agent. You can run it on your MacBook like I do at home. I love doing that. We can host it in a VPS. It's a powerful personal agent and you kind of look after it. Hyper agent itself runs in the cloud which is different than Claude. So Claude, you do it at your actual laptop. Hyper code, you don't really hyper code. Hyper agent. Oh my gosh, how many time how many coffees have I had today? Hyper agent runs in the cloud, which is really cool. Um, and it's something that it's easy for the whole team to message. So I could do it, my team could do it. Really freaking cool. Multiple agents, super brain, just to help you contextualize it against the tools that we talk about a lot and how it kind of fits into the universe. Another thing that's worth knowing here guys is on the um learning section, it will actually find improvements for you. One thing that's good to know, this is like a new technology, which is cool. And I've been pretty impressed by what they've done here. Your credits can build up uh pretty quickly if you're just sending it around in circles. So, just try to be judicious with the questions that you actually ask it. It can go off into the world and really think because obviously they've added some kind of prompts around this, which I think is really cool and helpful, which is why it was able to get the quality that they got the first time. But the upshot of that sometimes is that the cost can also go up. So, just be mindful of the models you're using and keep a tab on your car that's reasonable. But you do get good transparency in if you're having recurring agents what those prices look like. But again, it' be $500 down below for the first 500 people. Um, so feel free to grab that and have a little play around with it. And so now we've covered the three levels of AI agents. The next thing that we need to do is probably the most important step that you're going to make this year. And that's getting an agentic operating system for your entire life and your AI agents, which we learn how to do in this video right