No-Code AI Agents: Build & Launch Smart Tools with ChatGPT & SmythOS

1. Introduction : Hey, there. Welcome to No code AI Agents Build and Launch Smart Tools with Chachipit and Smith West. My name is Victor. I'm a digital content creator at YouTube and a huge fan of using AI to save time, earn more and automate the boring stuff. If you've ever wanted to build your own AI agent, but the word coding made you run for the hell you're in the right place. In this class, I'm going to show you step by step how to build an AI agent from scratch without writing a single line of code. We'll start from the very beginning by exploring the basics such as what even is an AI agent. We'll look at why they're more than just chatbots and how to plan your agent with Chat GBT in minutes. Then we're going to move on and build a fully working agent together using a free tool called SmithOS. Don't worry. If you don't use Smith OS, the concepts we'll be covering apply to any similar tool. You'll watch, as I create a real life AI system that writes LinkedIn posts, something you can use for yourself, your clients, or even sell as a service. By the end of this class, you'll be able to plan out your own AI agent idea, build it visually with no technical skills, and deploy it either privately or publicly as a custom GPT. I've broken this class into quick, by size lessons. So whether you're a content creator, a coach or just curious about AI, you'll be able to follow along easily. Your class project, you ask, you'll create your very own AI LinkedIn post generator and share your version with the class. Don't worry. I'll guide you through every step, and don't worry if you don't yet have access to a similar tool to what I'm using. You can also share a diagram, a mind map, theory of how you would build build your AI agent and share that with the class. As well, all the tools we're using are beginner friendly, mostly free at the time of recording this video. I've even included bonus pumps and templates to help you get the results faster. So what are waiting for? Let's jump in and build something amazing with AI. I'll see you in the first lesson. 2. Lesson 1 What Is an AI Agent : What is an AI agent? What are the use cases for AI agents? Could they apply to what you do in your life or are they just something that just wouldn't work with the way you work? Spoil our alert, whether you're a SAS founder, content creator, eComm business manager, C suite, CEO or executive agency owner. Chances are, AI agents have got a use case that will make your life. Easier. We go to cover that and a whole lot more. So whether you're new to AI agents or a seasoned veteran, there may be areas that you benefit from. The principles and the core foundations I'm going to be talking about here can be applied to virtually any AI agent platform. Let's start by understanding what AI agents are exactly. So obviously, they're a visual dragon drop platform that empowers users to build AI agents. These agents are intelligent, workflows driven by hat GPT or other models, various tools and APIs, unlike simply interacting with hat GPT in a chat interface. SmithOS allows for the creation of automated and repeatable processes. So in my experience, whether you're a beginner, whether you are a seasoned veteran, the learning curve is not as steep as I thought because they provide both templates as well as the opportunity to create AI agents from scratch. And either or, whichever option you go with, you don't need to be a developer. You don't need to be someone who writes code, you don't need to a professional with ten years of experience using AI, literally, you can do this using natural language and common sense and a bit of research if you are eager to learn and understand what all of this is all about. With a lot of these platforms, you may find it is rather challenging. There's a steep learning curve, but using this platform, for me, as someone who hasn't that technical of a background. With just a little bit of work, a little bit of research, I was building AI agents in next to No time. The platform is designed to be accessible to users without coding experience, enabling them to construct logic, integrate tools, and define actions through a visual control panel. 3. Lesson 2 Anatomy of a No Code AI Agent : So I got here a agent building cycle diagram that I have created here, which you can go ahead and check out from within the document, just a high level overview of what we are looking at here. The top step is building the logic. This is where you create agent logic visually. The way I do this and the way we're going to cover it in this video, I do this by interacting with ChatCPT and just using the platform to help me brainstorm articulate what I want to do in a way that not only Chuck GPT will understand, in a way that Smith Os will understand as well. Essentially, Smith O Wes works in a way where they've got their own master agent, which is called Agent Weaver, and this is their agent that helps you build other agents. So you want to be able to talk in a language that agent Weaver is going to understand. And that's where, for me, Chuck GPT really comes into its own because it helps me articulate all of that. So building the logic. And then we've got connecting the tools there, integrate with various tools and API. So depending on what your use case is, some use cases are really complex. You can make this as complex or as easy as you want. But some use cases are fairly simple. I don't have a highly technical background, so I tend to stay away. Super complex use cases need to access information from closed platforms like Facebook, YouTube, and TikTok. For context, what I mean by closed platforms, I mean that the agent is going to need to use something called an API in order to access that platform. And a lot of these platforms generally will sell an API API is like a phone call that your agent will need to make every time you ask for information. Once it gets that information, it displays it wherever you want it. The problem is because information is so valuable, a lot of these platforms will either not provide an API or if they do, it'll be a chargeable API. So if you look at it this way, the more valuable and higher quality information you get, the more expensive that API will be. You do get some platforms that are releasing completely free information that is up to date. However, a lot of the time, you will have to pay for these APIs. But the workaround found is with a lot of AI agents that I'm building, I tend to stick with information that is available via large language models. Smith of West, as well as other similar platforms will tend to a native connection with large language models, like Chachi PT, Claude Gemini, that in itself is a lake, a sea of information you potentially use to run an application, for example, writing scripts for TikTok videos, YouTube, long form videos, writing hooks, call to actions, creating product descriptions, and so forth. Generally, that structure that information doesn't change a whole lot over time. It tends to be with NSA even though the products themselves do evolve and change. So use cases like that, doing assignments for school work helping teachers write documents, write white papers and so forth. You could use a large language model, and a lot of these large models do tend to have either low cost or free APIs that are available in a nutshell. All this means in the simplest possible language is using LLMs as your source of information tends to be a lot less complicated than using APIs. APIs are really good as well, but there's a cost element, as well as a technical knowledge element that comes with specify actions. So you've got access to the API or various tools, specify what actions your agent is going to take. So you define agent actions through controls. I've built you as an agent. Here are the tools I'm going to need you to use. Now this is what I'm going to need you to use these tools for. Then you test and refine. This is where you evaluate and improve the agent performance. This is where you check for any bugs. This is where you're chatting with the agent, which I will show you a little bit later on in this video and seeing what type of output is coming out of the agent. So how a Smith OS agent works. So here's a simple breakdown of that. A Smith OS agent can be understood as a combination of brain and skills. The agent is the brain and the skills, an agent serves as the central intelligence responsible for determining what actions to take and when to execute them, the skills represent the agent's capabilities akin to arms, legs, and tools enabling it to perform specific tasks. So agent is brain and the arms and the legs and the body like the skills that you are going to give it. You can look at the skills similar to these tools that we're talking about here. Building an agent involves connecting these skills which are modular blocks of logic, data retrieval, formatting or tool utilization. I will show you this a little bit later on in the video so you can actually see it visually. Better understand the functionality of Smith OS. Let's explore its key components with relatable analogies. So this is what you're going to find as you're using the platform. Now, these are native to Smith OS itself. But if you're using another platform like an eight or similar tool, you will find a representation of these types of components in one form or another. But this is how they come across within a Smith specifically. You've got the agent, which is the entire workflow brain, skills, the little modular task inside the agent. As we explained, the agent is like a virtual employee. The skills are like apps or mini tasks that the agent undertakes. We've got inputs, what data goes into a skill. So what question Are you or the user of that agent going to be entering into a form or into chat, whatever it is that you're using for your website or your chat booard and the outputs, what data comes out of a skill? So answers from each task. Now, memory stores info across steps like a sticky note, the agent can refer to later. Not only does it give you the information that you asked for, it can come back to that information later on as well, tools, external services like Gmail notion, API is like giving your agent a phone or a computer, similar to making phone calls. Triggers. What starts the agent is a manual, an automatic trigger. This is like pressing a button or timer to get the agent started. Hopefully, that all make sense. So far, let me know in the comments. If you have got any questions specifically about AI agents, I will try and answer those as soon as I can provided that they are within the first 24 hours of posting this video. Terms of how the workflow looks, we have this representation here. So just briefly go through that. We initiate the agent. So you start the workflow process. This would be me or you entering a question or a request using the form the agent comes with. So interacting with the agent, execute skills, perform modular tasks. This is the agent now using the skills. You've given it the API or whatever it is, and performing the task that is required. It then processes inputs, so it gathers all the necessary data, generates outputs from what it's found online and stores the memory, so it saves the data in case it needs to come back to in future. And then from there, yeah, utilize tools, access external services, and so forth self explanative there. Consider an agent designed to assist in writing a video script, for example. So the trigger, in this instance, would be the user inputs a product name and description. I want to write a product description for these Apple earpotsF example. The agent utilizes a ChatPT model to generate a TikTok script idea based on the input. These platforms will have large language models like hathPT Built in natively, so you can easily access those without necessarily having to go and find an API key somewhere and key that in. Skill two, the agent refines the script to match a specific style or tone, I E, Victor's tone or someone famous tone that you'd like to borrow from Skill three, the agent delivers the final script by either saving it to a tool like Notion or sending it via email. Step four, the agent stores past prompts to avoid generating duplicate ideas in the future. You can see the assembly line of the workflow and the process that agents will tend to follow. Again, not just Smith or West agents, agents in general. Here are some examples of skills that you might use. We've got a variety of different skills here. Again, this might be platform specific. However, you may find similar variables in other AI agent platforms out there. So you've got Open AI chat for writing, summarizing, and researching tasks, Google search for retrieving real time data from the web, API call for interacting with external services like allo data or Shopify, conditional logic for implementing decision making processes. If this happens, then do that type of language. Formata for cleaning and manipulating text data, sending an email for delivering the final results of the agents work. So some examples of skills there. Now, what makes it powerful? So Smith OS offers several key advantages specifically. So modularity, skills can be built once and reused across multiple agents, promoting efficiency and consistency. Now, this is if you build these skills manually. I'm not I wouldn't say I'm at the stage where I'm confident enough to manually build these skills. However, if you are at that level and you understand a little bit more about the technical side of things, then if you have built that skill from scratch and put it together within Smith OS, which I will show you again in just a little bit how that visually looks, it can be used for a similar use case in future. They won't have to build it over and over multi step logic, agents can make decisions and adapt their behavior based on intermediate results. Tolsck integration similarly connect with a wide range of external services, CRMs, video tools, and forms. You'll be connecting via APIs. Collaboration ready. Agents can be easily shared with clients or team members facilitating collaboration and knowledge sharing. So building powerful AI agents, of got a visual here that we have put together. This visualizes what we've just spoken about when building agents for content creation or ecommerce is recommended to start with a simple design, one trigger, one GPT skill, and one output, you can then add layers of complexity by incorporating memory, conditional logic, and integrations with tools like stripe, TyfOm and notion. You can make it as simple or as complex as you'd like. For example, you could have multiple skills. You could have multiple different blocks of logic within the assembly line. But just remember, the more complex the AI agent is, the harder or the more work that it has to do. Maintenance might be a bit spotty as well, because if you have got issues, bugs yes, they can be fixed, but it might take longer to spot exactly where those errors are. Whereas, if you've only got three or four components, not only is it easier to identify what the issue is, but it's easier for Agent Weaver, which again, is like your manager, your assistant within Smith OS that helps you fix these types of issues. So really, the simpler, the better at the start, 4. Lesson 3 Plan Your Agent with ChatGPT : And then as you get more confident, you can add towards those. So, step one, in terms of how I built my agent, the agent I have built is a LinkedIn ghost writer, AI agent. You can see it here. Smith OS has this thing where they'll add a random Avatar picture for you. So this is not someone that's real, even though it may look. So it's called LinkedIn Ghost Writer for founders. Very creative name, I know. And here on the right hand side, we've got a chat window that allows me to test my AI agent by speaking to it. We'll get to this here in just a second. But what I've done is click on Test, and what I need to do now is just type the message to see what output my Air agent is going to. But moving quickly, let's step back a look and go back to our document, which is here. Step one in building an AI agent here. So for me, it was brainstorming with hatGPT. You'll find hatGPT or similar LLMs invariable when it comes to putting these types of AI agents together, especially for me, as I've noticed within Smith OS. And the reason for that is, again, hATTPT is able to articulate my requirement in a way that Smith OS is going to understand, is able to explain it in a way that the system is going to immediately go, Okay, this is what Vitor wants. These are the components. These are the platforms I need to connect to. This is where I need to scrape my data. I can't articulate that from my head. It's something I'm simply not able to do, which a lot of people won't be able to do unless you've been working with this type of technology for quite some time. So I start with hatGPT. The first thing I did is give Chat chip a prompt. I said, I want to create a LinkedIn post generator agent with Smith OS, ideally, only using LLMs already connected natively within the platform. These are text posts like blog articles for agency owners, coaches, SAS founders, and consultants. What do you need to know to draft me the best prompt from Smith OS agent Weaver? I want a LinkedIn post generator a come up with the idea for presenting and showcasing the technology within this video. It can be your idea, whatever it is that you want to create. However it is, you've gone about doing the research for creating that. I'm also telling ChatiPT that I only want to use large language models within Smith OS. The reason behind this, as mentioned a little bit earlier in this video, I'm trying to keep it as simple as possible. Smith OS could very easily make this very complex if I'm not specific about where I want the information to go. Could think of ten different platforms, APIs, freely available APIs online that they can connect to various different sources. So if you don't tell it, I want to use hat GPT specifically. I want to use that skill with open AIs chat because I know that's already available within the platforms. The workflow is going to be so much simpler, as opposed to letting it scour the deepest darkest corners of the Internet and just come back with errors. I'm also specifying that these are text posts like blog articles or agency owners and who it's for. Now, the key part, and this is something that I repeat over and over again. What works for me when it comes to chatting with ChachiPT or these similar platforms, rather than giving it a load of information, I tend to ask it in order for you to get me the best possible result. What do you need from me? Remember, Chachi Bit is already 100,000 times ahead of you in whatever it is you are trying to do as far as information access is concerned. So rather than saying, Hey, go do this, I'm like, what do you need from me in order to get me the best result? He then turns into that expert and says, Okay, you need to give me one, two, three, four, five, six, seven, eight, and once you've done that, I'm then going to create what's needed, and we can move on from there. So this is what Chachi Pit came back to me with true to form, just like I've explained. It said, give me an ideal audience or personas, who exactly shoot these posts appeal to. Give me a style and voice, some examples there of what he's looking for. What are the content pillars that we're going to write about or categories, main business goal, output details, optional extras, blah, blah, blah. So this is, again, ChachiPT just kind of coaching me saying, Once you tell me this, I will give you the best possible result that I can. So I went ahead and answered ChachiPT's questions, which again, very simple. I've just used some of what it's given me to answer the questions back. Again, you can make this as in depth as you want, but for the sake of this presentation, I am just giving high level information based on the information that it's given me there. That's me answering hat GPT, right there. And in terms of this chat, you can do this within the free chat GPT platform, but I'm using the paid version because I just find it so much more useful and so much more helpful for my use cases. 5. Lesson 4 Build Your Agent in SmythOS : So I gave the information back to Chat CPT. We then moved on to step two, which is pasting the hathPT prompt into SmithowS and creating the agent. So this is the prompt that ChahPT came back to me with after I gave the information. Now, for context, this prompt is what I'm going to paste into Agent Weaver. You can see on the screen there on the left hand side, you can see Agent Weaver. This is that master agent I was talking about at the beginning of this video. This is the manager within this platform. It articulates what it needs, and it does all the running around required to build these components here that it went ahead and build for me. So what I did is take the prompt that Chachi PT created. You are a seasoned LinkedIn course writer for agency owners, blah, blah, blah, it's now talking to Smith OS. With this prompt, follow this structure for every post, hook, bold curiosity driven, first sentence, story lesson, CTA, tone, randomly alternate topics across two categories. I do not include hash tags or emojis. I've asked you to post one post per run, and then from there, yeah, we have got our prompt. So I've taken this prompt, and I have pasted it onto Agent Weaver, as you can see there. And what agent was done is then go ahead and let me know. I'm going to create a Linked in ghostwriter agent that generates emotionally resonant blah blah blah. So it's just repeating what I've put in there in summary, and this is it working here. So it went ahead and created this for me. For context, the reason why I'm not doing this live, I respect your time. I'm not going to sit here and let you wait minutes while it's creating this agent. We can do this live another time, but I just wanted to walk you through what I've already built. It doesn't take that long to build it, but I just didn't want to do that. So yeah, as you can see, it's letting me know. It's thought process as it's working here. And there we go. The agent is ready to generate high quality LinkedIn posts that position your client as thought leaders. So while Agent Weaver's going through this process, it's creating these different components that you can see here. Let me just close this window to give you a closer look. Just going to zoom in. So that's the agent itself. As you can see, the skills that we described earlier, this is where you can add any new skills if you wish to. I have not done that. I've let this agent build the current ones for so we've got this first one here, which is an API endpoint. We talked about APIs a bit earlier in this video. It generates a LinkedIn post for founders and coaches inputs are topic category, newsletter info, and we've got all this stuff here. We've got the AI LLM next here. The source of the information is Chat GPT, which is the main LLM connected to Smith OS. Once it's collected, this information is going to post that output. As you can see there, that's a LinkedIn post. You could be a ghost writer, C suite, CEO, director, whatever use case is, you could be a consultant coach. But from there, the tool is ready. So with that being created, the next step is to test the agent output 6. Lesson 5 Test & Refine Your Agent : I've not really put down the steps, but all you do when you test it is you just go onto this test button here on the top right hand side, it's letting me know who it is and what it does. If you've watched one of my previous videos, you will know this. I tend to ask it, what does it need from me to give him the best possible results similar to what I do with Chat TPT, so I'm going to do the same thing here. Boom. Okay, so to create the most effective and resonant Linked in post for you, here's what would help me deliver the best results. Target audience, your primary goal, your personal backstory, tone values, and speaking that language that is going to help you get me the best result here. So because of my suffering from chronic laziness, I'm going to copy this here. Perfect. Let me just copy all of it. And I'm going to put it into my hat GEPT chat. And I'm going to ask Chat ChiPT to generate a dataset for me with just some fake information, answering all those questions, I can copy and paste back into that air agent there. I'm just going to paste it here, go to the top some example information ering the following request so I can copy and paste back to voice generator to get a result. I'm going to let Chat ChiPT think in here. So here's a fully filled out example, response. Target need to wait for that. If you are finding value a relatively small creator, more people can see this video, and the more people like a video, the better the response tells me that I need to create more of that content and even consider subscribing if you feel that I deserve it. There we go. Target audience. I've got everything I need here from chat GPT, so I'm going to copy that and go back into the agent, give it the answer. There we go. It's happy with that. It's now writing the post. Let me know how it's going to approach it. Okay, you can see now it's working through the different skills. It's looking at the moment through the Gen AI LL. It's got the information that it requires. Now it's writing the result. I've only asked it to create 150 to 300 words, so it's not going to be a long post. Can AI be the partner who doesn't take away your authenticity, blah, blah blah. Not long ago. Having seen it. I took a beautiful. So that's ready. All I can do now is just copy this and paste it into LinkedIn, and I've already got my posts for the day I can tick off another to do task. Or if you're a ghost writer and you've got five clients, I expect you to post every single day, enrich that, copy it, chop it, change it, whatever you want to do. 7. Lesson 6 Deploy Your Agent : But you've got 80% of your work done right there. Now, for the sake of time, I've tested the agent output now. The next step is to deploy the agent. I'm not going to go too in depth in deploying the AI agent here, but what you do is you click on this deploy button. If you encounter any issues, the best approach for a beginner is to interact with Agent Weaver. Agent Weaver is again going to do all the running around for and identify, sniff out the issue, and a lot of the time is going to point out exactly what the issue is, and it's either resolve it or suggest an alternative. A big one that tends to come up is if it's unable to scrape information based on a source that you've told it to go and get the information from, then it asks you, do you want it to change the source of information from Shopify to WooCommerce, just so it can get a better source and get the information that require, things like that. Just do a natural language conversation with this, and you can fix any issues. But yeah, once you do, if you're happy with the agent, if you're happy with everything here, you can go ahead and click on Deploy. You can see different options. You can do it as a domain where you put it onto your website directly or your application directly, or you can deploy it as a custom GPT, which is my favorite option because it's the least complex option. You can deploy it as a chatbot to put onto your website, as well, API to connect to other applications. If you want to act as a white label background application like Smith the West themselves. Just a very high level to deploy as a custom GPT. You just click on Get code. You've got the Dev uRL there, copy that, and then you just click on here. If you've got a paid version of Chat GPT, which I do, is going to take you straight onto this new GPT screen, and this is where you can create and configure. The agent, you can add picture, the name of your agent, the description, any specific instructions you'd like it to follow, while it functions, conversation starters. We can add knowledge here and some additional capabilities. I would advise not to because it's got all the capabilities that it needs right here. If that's something that you want to do, you can add any additional actions that you'd like it to take. And this is where you can test the GPT. You can see here already what it's going to look like. And yeah, that's what you would be doing there. And once you're done, you just click on Create. And just like that, you've got your own custom GPT, which you can either sell or use as a lead magnet to get people to sign up to your newsletter, your service, whatever it is that you 8. Conclusion + Next Steps : And that's a rap. You've just learned how to go from zero to agent planning, building, and even deploying your very own AI tool with no code, no tech, headaches, and a whole lot of possibility. Whether you followed along and built the LinkedIn Ghostwriter agent with me, or you're now thinking about what kind of AI agent you want to build next, you've officially got the tools, the process, and the confidence to do it. Now it's time to put it into action. Your class project is simple. Build your own AI agent using Smith OS or a similar tool and share a quick screenshot or short description of what it does within the project gallery. And again, if you don't have access to such a tool, don't worry. A diagram or mindmap displaying your theory for an AI agent will suffice as well. Even if it's not perfect yet, share it. I'll be checking in, leaving feedback, and celebrating your wins. If you've got questions, ideas, or want to connect, drop me a comment below or find me here on Skillshare. I'd love to hear what you're building. With AI. Thank you again for taking this class. It really means a lot, especially if you're one of those people who thought this stuff was too technical before. You've proved it isn't. You just needed the right guide. I'll see you in the next one.