This is the full transcript of How to Build an AI Agent to Analyze Facebook Ads Using Cursor, Claude & MCPs, published on YouTube by The Boring Marketer. Every paragraph carries the moment it was spoken, so you can click any line to jump straight to that point in the video, search the whole thing for a word, or copy it out.
0:00Okay, I just started learning how to use Cursor for Vibe Marketing, and so far I'm totally blown away by the power of this tool. So, there's no need with Cursor to build complicated workflows with lots of different nodes and things like that. You can actually do everything that you need with the agent within Cursor and execute Vibe Marketing
0:24strategies right from the command line. Now, I'm not a developer. I just started using cursor about I don't know three days ago and every day I've been doing a small experiment to try to get familiar with this platform because I think it's going to be such a powerful tool for people to use that aren't just developers, marketers, operations, things like that. Anything you want to automate or build, you can do it right from here. So, I'm going to give you a view of the output of this experiment that I did and I'm going to walk you through it. So, as a marketer, I have been running a lot of ads on Meta and
1:05things like that over the years. I still have a an ads agency that is doing this kind of work. One of the things that we do is we go and try to look at competitive ads that companies are running and analyze those and get some insights. Now this was a pretty manual process uh in the past. So I wanted to get some ideas on how we can do this programmatically using AI using cursor leveraging MCPS and all of that fun stuff. So the tool stack that I use to build this analysis I've got cursor here and for my agent I'm using the Claude 4.0 model. It's super smart, very good
1:47at coding. I've been thoroughly impressed with Claude's latest release so far. I'm using Appify, which has a scraper that's going to go out to uh Facebook pages or uh brand pages and it's going to scrape those ads and help me store them in cursor for analysis. And then I'm using the playwright MCP. Now, playright is a browser operator that I can call right from cursor once I have that MCP installed. I'm going to show you how it works. Don't worry. Uh, but that's the steps that I went through. The goal was to build like a comprehensive analysis of Facebook ads from brand to scraping to actually
2:33having a playwright view the ads for me, whether they're videos or images, and then derive insights. What are the patterns that these brands are using? What kind of visual styles are they using specifically for the longest running ads? Which is a bit of an indicator to me that perhaps these are the top performing ads. So, you can do this for any brand that you want. Here I've got a bunch of random car dealerships just to do a fun example, but let me give you a uh a view of what this looks like here. So, again, these
3:06have actually been viewed by an agent. the playwright MCP and uh automatically cursor has built some code to give me the pattern. So let's go through it a little bit. Um it analyzed 25 ads that it actually went and viewed for me. I didn't have to click a bunch of links and do it myself. And I'm doing all of this from cursor. Uh it gave me an executive summary. So, uh, what are the patterns between the longest running ads for this particular niche? Uh, it's got minimalist brand focus, lifestyle positioning, and a consistent visual identity. Uh, and then going through each and every brand, it tells me what
3:50it observed and what the overall creative strategy of this brand is. So, for this one, for example, it's saying it has premium lifestyle positioning. It's giving me the elements that it saw when it actually went and clicked on those media links and the scrape file. And then it's giving me tactical insights. So, I've got this for a multitude of brands. And then, uh, it's even giving me like key performance indicators. Um, what are the highest performing creative elements that are, uh, happening between these longest running ads? So, a square format, lifestyle photography, minimal text overlay, consistent brand color application, professional lighting. So, to do this for hundreds of ads or
4:36whatever you want would take you such a long time manually and there would be a lot of sort of like inconsistent insights and and things like that. So, I think this is incredible that you can do all of this from cursor. Um, again, there's not a lot of nodes. There's not a lot of different things and output files and stuff like that that I have to go and access. Um, you know, tactical creative recommendations, brand creative scoring, uh, creative excellence ranking, um, just incredible stuff. So, just to give you, uh, again a quick view in terms of how I set this up. Um, so if you use the Appify Curious Coder uh
5:18Facebook ad scraper, you'll get an output file some that looks something like this. I instructed the agent to to simplify it. You're seeing the simplified file. The most important thing that you need to collect from the scraping output are these media links right here. The media links are what you're going to instruct your playwright MCP to actually use your browser and go and look at and analyze and then bring that back into your cursor environment and you can instruct the agent to you know summarize the insights and build you a markdown file like this. So this is uh this is pretty wild. Uh again, three days of uh cursor testing and
6:05essentially I've built um a Facebook ads agent that can go out, find competitive ads, actually look at the images, watch the videos, and give me insights. I think there's a lot more that I can do with this, and really just feel like I'm scratching the surface with this platform. Uh I think that the progression of Vibe marketing goes from using siloed tools and siloed LLMs. You know, you might be going to Chat GPT to do something. You might be using some sort of AI tool to go do something. And then the next step is you're starting to chain those actions together with a workflow builder. Um you know, maybe
6:47that's inate, maybe that's gum loop, maybe that's plum, something along those lines. After you're able to sort of understand the logic and chain things together, uh the next step is starting to learn how to work with MCPs. So Claude is a good step in order to to do that. But cursor is much more powerful in the sense that you can access
7:09multiple LLM models in one interface. You can access a ton of MCPs and set them up pretty easily without having to set up nodes and inade or whatever that looks like. Uh and then you can code, you can create things, you can derive insights. I'm not even using cursor as something where I'm deploying production apps or anything like that. I'm just using it as my analysis tool, my automation tool. Um I can create content with this. I can analyze my competitors with this. Uh it's wild stuff. I hope you get some value out of this. I'd recommend checking Cursor out if you haven't yet. uh they don't pay me to to
7:51talk about them or anything like that. I'm just kind of blown away at the vibe marketing use cases uh for this platform after the video I did uh a few days back. So give it a shot. I think you'll be uh you'll be surprised as to what you can do within Cursor. Good luck.
No line in this video contains that word.
Where these words come from. This is the caption track YouTube holds for this video, written automatically by YouTube rather than by the creator. We read it, tidied the line breaks and laid it out so it can be read. The plain text version is at https://viewrankai.com/tools/youtube-transcript/pQ5NqM2NXqA.txt.
All rights in this video belong to The Boring Marketer. Watch it on YouTube. If this is your video and you would rather this page did not exist, tell us and we will remove it.