Build your first AI agent in under 23 minutes (Cursor, no code)

Instead of paying $200 a month for Chad GPT Pro or hundreds a month for Manis, you should just build your own AI agent for free. In this video, I'm going to show you how to build your own AI agent that goes on the internet, does work for you, and builds you research report. It does it all completely autonomously. It's like your own AI employee. Instead of paying for deep research, you can now just build your own AI agent that does deep research. Check out what we're going to build right here. You're able to go in, tell exactly what you want your AI agent to do, and it's going to go on the internet and do work for you. So, I can give it a mission that it's going to accomplish online. I can say, "Start the research. " This is something people pay hundreds of dollars for OpenAI or Manis' agent. I'm going to show you how to build it for free. You give it a mission. It takes the mission, breaks your mission down into multiple steps, and then when you say execute, the AI agent goes on the internet and starts executing on each of these steps so that it can build you a full research report. It then goes takes the stepby-step plan and start it starts executing on it online for you on the internet doing research, completing tasks, and basically working for you. And this is done completely autonomously. If you lock in with me here, by the end of this video, you will have built your own AI agent that goes on the internet and does work for you. Let's lock in and get into it. Real

quick, before we start building in cursor, I want to talk about what makes an AI agent an agent. So, you know exactly what we're about to get into. There are five things that we are going to build into this AI agent that makes it an AI agent. This is the difference between like an AI chatbot when you go into chat GPT and actually interacting with like your own AI employee that does things for you. Those five things are planning. So when you give it a goal, it actually builds a plan to complete that goal step by step. Tool use, it has a set of tools it can use to execute on that plan. Autonomous, it executes on that plan autonomously. So you can sit back, walk away, and it does its own thing to complete that plan. It's reflective. So as it goes, it looks back on what it did, reflects, and can improve its plan moving forward. And then goal oriented. So you're giving it a goal, and that's what's driving everything you see here. It h it builds a plan to complete that goal. These five things are typically things just straight up AI chatbots don't have. This is what we're going to build right now into our AI agent that will make it our own AI employee. And don't worry if you don't understand some of this. We're going to make this dead simple. You're not writing any code. You don't need to be technical at all. We're using AI to build all of this out. an AI agent. A little dystopian, but we're doing it anyway. So, where I like to start off is in Claude before I

start building anything in cursor. I start off in Claude just to get kind of my initial prompt into curs. That initial prompt to use to start building things in cursor, which don't worry if you're not used to cursor. I'm going to show you how to use all this in this video. This is built for beginners. The thing is you want that first prompt in cursor to be powerful. So, I use Claude to actually build that prompt. So, here is the prompt I'm going to use to give to Claude to get the initial prompt into cursor. I'm going to put this down below in the description. So, feel free to copy and paste this into Claude as well and follow along with me here. Hi Claude, I want to build a research AI agent in cursor that I can give a mission to and it completes the mission for me by breaking it down into steps. Then completing those steps. It then has access to web search via Tavi to gather information for that task. I want to use Nex. js to build this agent inside cursor. Please provide me initial prompt I can use to give cursor to start building this agent. So this is the initial prompt. It tells us a high level what we want to do inside cursor. And then Claude is going to build an initial prompt we can put into cursor to start building that agent out. Claude's going to make it so that prompt is super detailed so we get a really good first step when building this in cursor. So let's hit send on this. Okay, looks like we got the initial prompt here. build a research AI agent with Nex. js step one. Okay, so it's making clear we want to do this step by step. It describes what we want to get to eventually, which is an AI research agent that can take any topic, do research, and create documentation around it, and it goes through development approach, goes through a tech stack. This is looking good. We're going to take this initial prompt, and we're going to put it into Cursor in a second. And for those new to the channel, we use Cursor and Windsurf to build out our applications. For this one, we're going to be using Cursor. There's a free tier, very generous free tier. You can download it now. Cursor. com. Go there, download it. You're good to go. This is our AI that's going to be building the app out for us. Makes it very, very simple to build out apps. Cursor. com. All right. So, we are

inside Cursor. Over on the right is the cursor composer. This is your AI agent that is going to go out and start building your application for you. We're going to be using that to build out our entire application without writing any code whatsoever. It's going to be super simple. And at the end of this, stick around with me. We're going to have an entire AI agent that goes out and does work for us on the internet, which is amazing. If you learned anything up to this point, make sure you hit the subscribe button down below. Make sure you leave a like. All I do is make awesome videos showing you how to build stuff with AI. All right, so we're in cursor. Make sure to select agent down here. This makes it so that it goes out and builds code for you. I'm using Claude for sonnet for this, but you can use really any model you want. Claw for sonnet does cost money, but you can use claude 3. 5 as well. that is significantly cheaper and it'll give you similar results if we're being quite honest. All right, I pasted in our initial prompt that Claude gave us here which goes step by step on exactly what we're going to be building and I'm going to hit enter on this and it's going to start building the application for us. All right, it's going to help us create a clean Nex. js project. And for those who don't know at home, Nex. js is a web application technology used to build some of the most popular web apps out there right now. we'll be using for this. Very simple and easy to use for the AI agents that build out our applications. So, let's run this command that installs the project for us. We hit run. Very easy. Cursor runs the commands for us. It writes the code for us. It really makes you don't need to be technical whatsoever. So, it's building out the application. It's installing all the basic technologies we need. Now, that's been created successfully. Let's navigate into the project directory. Okay, let's do that. I hit run. It goes into the project. That's awesome. Now, let me check the current app structure to create the additional. So, it's setting up the files. directories for us. It's doing everything that would usually take hours to do, it's doing for us in seconds, which is awesome. Okay, so it ran npm rundev, which runs our application for us. Let's see if our base demo application is built before we start building the AI agent on top of it. So, we'll go here. We'll go to localhost 3001 research AI agent. Our base application is set up. We are ready to build, baby. We're ready to build. Our web app is running. If you never built a web app before, congratulations. You just built your first web app. Now, we're going to start building out the AI agent on top of this. What's great about using the prompt that Claude gave us is typically when building applications in cursor, you need to kind of figure it out step by yourself, right? You have to figure out what do I need to build next. And that sometimes takes some technical knowledge to do. But what's great about building a very in-depth prompt with Claude first and then putting it into cursor is what you can see here what's next. It already knows what to do next. We don't even need to guide it on what to do next. We just got to say go ahead build it for us right and this is what's great about having a co-pilot CEO AI chat up as you build like I showed you with Claude is it does all the thinking for you. We don't have to do much of the thinking ourselves. So that's why I always have another window open. Whatever the most powerful model is at the moment I have that open. Right now it's Claude sonnet 4, but whatever it is at the moment, you can have that up and anytime you need to build out prompts or have questions, you can ask the chatbot and then you can put it in the cursor. So now I can go, okay, that works great. Let's move on to step two, which is building out the input for the agent. And I can hit enter. And that now it's going to start building out the UI for the agent for us. All right. So as you can see here, it's writing the code. We don't have to do anything. We're just watching it write code for us. It's building out the interface. Uh and then Okay, so it's all done. Let's see what the interface looks like here. Oh, that's clean. Research AI assistant. Okay, so we can enter in a mission. It gives us example missions, which is sick. This is great. So I can put that in. I'm going to imagine when I hit start research, nothing happens because this is just the UI. But yeah, that's great. Okay, you can see what the processing looks like and everything. Mission received. Oh, so we got this screen where I can see the AI agent confirms it got the mission. All right, so now we need to actually start building out the logic. Let's review where we're at when it comes to building AI agents. Planning, tool, use, autonomous, reflective, and goal oriented. So we built out the goal oriented part, right? We have the ability to give it a goal. Next, we're

going to build out the planning part. So we can start building a plan to execute on that goal. And that's going to be done with some nice AI trickery here. So for the next prompt, when we want to start building out the prompting, we're going to ask the AI to do this. Let's just go straight into building the AI integration out. Let's integrate this with the Open AI API. So, we can send the mission to the API and it returns us back a step-by-step plan to execute on. So, we're going to be using the Open AI API, which basically means we can send our mission to chat GPT. Chat GPT will do processing and then send back our AI agent a step-by-step plan to execute on. I hit enter on that. Okay, so it is installing Open AI into our application, which is sick. It's going to create an environment variable file for us, which is great. And I'm going to walk you through in a second exactly how to get like things like your API key and things like that so that you can actually have the interactions with the AI. Everything here is built for beginners. So don't worry if you're not technical or any of this is confusing you. I promise I'll make this simple for you. All right. So now it's making folders for our new components and for the API calls. So for those who want a little bit more context on how this works, it's going to basically send messages to OpenAI saying, "Hey, here's a mission. can you build a plan for us and OpenAI will send a message back saying here's your plan that you can execute on which will then be displayed in our application. This is really cool also. It sets up a readme that describes everything the application does. So if we ever pass this code off to someone else, they'll be able to understand what's going on. All right, looks like it coded it out. Let's just create our environment variable. This is where we'll put our API key for OpenAI so it can make the calls. We go over here to the left. We rightclick. We do new file. We can call it enenv. local. Hit enter. Then we can take this code here. Copy paste openai API key. And then we'll put the API key here. I'm going to show you how to get the API key right now. If you go to platform. openai. com and then you click log in or sign up, you can get your API key here. So I'm going to sign once you're all signed up, you can go in. You can click API keys in the lefth hand side. You click create new secret key in the top right. Now we can create a secret key for our API. I'm just going to call this agent demo and I'm going to click create the secret key and that'll give us a key. Once you do that, paste that into your. env. local file and you're good to go. All right, so let's test this out. Let's see what we get. All right, so I'm going to ask what are the major differences between cloud 4 and claude 3. 7. This should send this to OpenAI and get us back a plan to do the research. Let's start the research. All right, so look at this. It has a research plan generated. All right, so it has uh gives us an overview of the research. That's sick. First of all, this UI is pretty amazing. This is a beautiful UI. All right, so research steps. Gather technical documentation, analyze performance benchmarks, conduct user surveys, review expert opinions, synthesize findings, prepare presentation, solicit feedback and results. Wow, it has an entire plan for us. That's pretty amazing. I want to make one small change here to this research plan with the AI. I want to make sure the AI agent knows the only tool available to it is web search. Some of these things request like interacting with users. The AI agent won't have access to human beings. It'll only have access to a web search tool, which you can add on more tools later in the future after this video if you want. Maybe I'll do a follow-up video to add more tools. But just for this video, we're going to do just web search. Let's make sure the AI agent knows it has only access to web search tools. So, I'm back in cursor. Anytime you want to make changes to your AI agent, you just go right over here, give it a prompt, it'll make the changes for you. All right, so I'm just going to send this prompt that says, "One small change. Make sure the AI is aware. It only has access to web search as a tool. " I hit enter on that. It's going to make sure the prompt that sends to OpenAI is aware of all the tools it has access to. All right, looks like it updated the prompting to the AI. Let's just test this one more time. Going to put this in. Start the research. Let's see if this plan is more aligned with the tools our AI agent will have available to us, which is just web search. All right. Search for recent articles. That's good. Analyze blog post. That's good. Find publicly available reports. Research social media. Compile. Okay. So, this is all web search relevant. I love it. This is a really good plan. So, we took care of goal oriented. planning. Now, let's make this so it has access to

tools. This is another keystone to AI agents. All AI agents must have access to tool use. So let's implement web search now for our AI agent so it can go on the web and start searching things for us. All right, that worked amazing. Now let's implement web search for our AI agent. So it actually has tools to use so it can actually go out on the web and do things for us. For the web search tool, we're going to use an app called Tavi. Tavi again very generous free tier everyone can use. I'll show you how to sign up for that in a second. But we're going to ask the AI to go in cursor and build out the Tavi integration so that our AI agent can then go take our planet built and then actually execute it on the internet. Let's do it. Hit enter on this. So I said that worked great. Let's move on to implementing Tavly so the AI agent has access to web search. So now it's going to start building it out. I hit run on all these commands. Starts running it for us. It's going to start building out the folder structure for us. I just hit run on that. And as this is running, I'm curious. Let me know in the replies what's going to be the first thing you have your AI agent research for you on the web. I'm curious to see what you guys are interested in right now. So, it's building out all the code for us. It's making sure Tavi is set up correct. In a second, I'll show you to get the API key for Tavi so you can have it search the web for you. I've built a few AI agents before and Tavly is definitely my favorite tool to use so that the agent can go and have access to the web. I absolutely love how it updates the readme automatically as it's building out. That makes it so easy to add on people to your team if you want to keep building this in the future. That's really cool. Okay, so now we just need to add the Tavi API key to our environment variable vial. So I'm just going to copy this and we're going to paste it in. Once you pasted that in, head over to tavile. com. That's tavv ly. com. Sign up. It'll give you an API key. It's free to use initially. I haven't even put in a credit card number. Then you can just copy and paste this into your environment variable file. All right, so we're back in the AI agent. Look, new button, test tavly search. Okay, so here's the plan. We're going to test the Tavly search out just to make sure it works. It's important as you're building this out. I mentioned this at the beginning. You want to break everything down step by step, right? You don't want to just oneshot everything. You break it down step by step. You build it out step by step. And every step, I'm saying step a lot. You're testing along the way. So, we just implemented Tavly. Let's test this step out. Let's test out the search. Once that's good to go, we'll connect the two from the researcher to the web search. So, the two can connect and our AI agent will be good to go. So, let's test this out. I click test. Okay. Electric vehicle market analys. Let's just take one of these example ones. I like that idea. Let's click that. Electric vehicle market analysis. Let's see how powerful Tavi is. Wow, that was the quickest search I've ever seen in my entire life. Response time 3 seconds. That's incredible. Uh AI summary. The electric vehicle market is growing rapidly driven by increasing demand and government incentives. Wow, this is actually unbelievable how it did this in 3 seconds and it gave me five sources live from the internet. That's pretty amazing. Electric vehicle market size. So, it found us five articles and then gave us an AI summary. And this again, this I really like the way this UI looks. So, it looks like the web search is working. Now, let's connect the two. Let's go back over here again. These are the five things you need to build an AI agent. We entered in the goal, so it's goal oriented. It does planning. We just did the tool use. Now we're going to

implement the autonomous and reflectiveness so that it works on its own and then it reflects on what it does and improves it. So let's connect the web search to the plan so it can work autonomously through the plan using its different tools. And then once all five of these are complete, we have our AI agent and then you can add on top and do a lot of really cool things. So we're back in cursor. We're going to talk to the AI agent here. This is the prompt we're going to use. This is looking good. Works great. I always like to compliment the AI so it knows what it's doing well. So it like reinforces itself. Let's connect the web search to the plan so that when we enter a mission, the agent makes the plan and then executes on it using Tavi. Hit enter on that. It's going to connect the web search to the plan building so that our AI agent can build its own plan and then start executing on it by itself. Excellent. Now let's build the exciting part, the automatic plan execution. So it's going to start writing the code for us by itself. Look how deep we are into an AI agent right now. and we haven't written a single line of code. That's pretty amazing. So, it's writing all the code for us. It's building out the component that takes the plan and then executes on it. So, it's really amazing what's happening here. The AI is building out our agent so that it does the step by step. It goes it builds the plan for us for what we're researching and then it goes it takes that plan automatically executes on it. Okay, so it's updating the plan component. All right, let's test this out. All right, looks like it's complete. Autonomous research AI agent complete. Let's see what we got here. So plan execution is got everything. Professional report layout, export functionality, research breakdown. Let's test this out. This is cool. So we have apparently we have our research agent built. Let's test this out here. npm rundev to run the application that runs it. Locost 3002 for me might be different for you. All right, let's see here. What is the difference between claude 4 and claude 3. 7? Let's start the research. Let's see what it does. Let's see how it works here. All right, so it's processing which means it is probably building the plan with open AI of what it's going to execute on and search. All right, so it has a plan here. Search for recent articles, analyze blog post, find publicly available report, research announcements, analyze social media, and compile the f uh the findings. And then down here, execute the research plan with AI. Let's do that. Now it's going to start executing it. Let's execute the plan. Oh, this is sick. Okay, so it has all the steps and now it's going to go through autonomously and complete each step using the tools it has available for us. For now, it is the Tavly web search tool. So, it's going in, it's going step by step, it's searching the web to get the information for each step, and it's going to compile it into a report for us. All right, so all eight, all six of them are done. Let's just see here. Let's view real quick the results. Okay. So, you can go in each one and you can view the results of what it searched for each. Okay. So, it gets a bunch of different articles for each steps. Goes through all of them, summarize them. All right. So, let's view the research report. Let's see how this does. Okay. So, here's the final research report and we're going to improve this. I'm not in love with the way this looks. So, we're going to improve this in a second here, but the final research report shows us all the steps. It went through the sources. So, it got 40 different sources. It has an executive summary. So, it has bits and pieces of the executive summary. I'm not in love with this executive summary because it doesn't it shows a summary of each step. I want an overall final research report that just gives me the answer. So, let's improve this in a second. Shows the key sources. I like that. And then the research breakdown. I like that. But let's improve this final research report. So, we get a really, really nice report. So, the prompt I'm going to give is I'm not in love with the final research report. I want to take it so we take all our sources, all our information, send it to OpenAI, and get back a complete report. I'm hitting enter on that. It's going to work for us. Again, the key here, we're just talking in plain English to the AI. You don't need to know coding. anything technical. We're just sending in plain English exactly what we want in our app. The AI goes and builds it out for us. So, it's building out the components to generate this very complex, in-depth final report for us. It's making the folders, the directories, the components. Again, I'm just sitting back. I'm just drinking my water and enjoying the show as the AI agent builds out our own AI agent for us. All right, looks like it's set to go. Let's try this out, shall we? All

right, let's give it to our AI research agent again. We're going to start the research here. We're going to see what the difference in claude 37 and 4 is. Start the research. Let's see if we get this very indepth research report from the AI agent, our employee that's going out doing research, building a plan for itself, autonomously, working, autonomously, doing research on the web for us. People are paying hundreds of dollars to do this with Chat GPT Pro right now. We're doing it for ourselves for free. All right. So, it built its plan. Let's execute the plan. And it is going. Let's see what we get. I want to see this final research report. Again, a lot of companies like Manis, for instance, are charging hundreds of dollars to do this a month. We just in the course of what 10, 15 minutes built this ourselves for free, which shows you the power now of what you can do as this is executing here. Like, if you have cursor and you can think of prompts to put out there, you can build anything you want. Anything that you're paying money for at the moment, you can just build yourself with cursor and just save yourself tons of money, which is incredible. It's a amazing time to be alive right now. All right, let's view this research report. Okay, looks like the research reports done. Let's check it out here. All right, executive summary gives me a highlevel overview. Claude 4 compromising of opus and sonnet versions represents a significant advancement. Here's the key findings. Oh, I love this. It breaks it down into seven key findings. Claude 4 opus and sonnet models offer superior coding and reasoning capabilities compared to 37 particularly in handling complex longunning tasks. Gives me a full breakdown and gives me a detailed analysis and then a conclusion. Wow. What is this recommendations organizations should consider integrating cloud 4 into their coding workflows to leverage its superior. That's amazing. Wow. So it gave me this entire full research breakdown. This would have if you use deep research from open AAI cost you $200 a month because you need open AI pro to do that. If you're using Manace you had to pay hundreds of dollars a month. We just built this for free. This is completely free. This research AI agent builds you this indepth report based on 40 different sources and it did it in 4 seconds. That is amazing. That is incredible. We built our own AI employee. So, anytime you're doing work and you need some information, you have a question, you need a big research report, and you need someone to go on the internet, find a bunch of different sources, and do research for you, this AI agent is that I showed you how to build it. This is the power of AI agents in the future is they're employees that work for you. Right now, you're able to go do work while you're building something out. Have your AI agents go and do research for you, find information, and just up your productivity a ton. This is a major trend for this year. And if you followed along and built along with me, you are ahead of that trend because you now have an AI agent that works for you. If you wanted to continue to evolve from here, you can add on different tools, right? If you wanted to make it so it can control your desktop, you can add that on. You would just work with cursor to build that out. If you learned anything at all, make sure to hit the subscribe button, leave a like, and I'll see you in the next