What is Firecrawl?

This episode is the clearest explanation of FireCrawl on the internet and how you can use it to build a real business that makes you real money. FireCrawl feels like giving your AI eyes. Right now, AI is smart, but it's blind. It can't see the internet, it can't go to a website, it can't grab data. So FireCrawl fixes that. Once you see it in action, it changes how you think about building products, how you think about collecting data, and how you think about what's possible with AI. In this episode, I break down what FireCrawl actually is, how it plays into your AI stack, and walk you through a bunch of startup ideas that you can make money from it. I use FireCrawl with ideabrowser.com, and I reach out to them to ask them to sponsor this video. They said yes, so that more people can see this, get the sauce, and build and make money with it. If FireCrawl has been on your radar and you just want a clear explanation of what it is and how you can use it as a founder, then this episode is for you. And if you've never heard of it, honestly, that's even better. Because what I'm about to show you is going to change how you think about what you can build with AI and where the next 12 months of building is going. Let's get into it. By the end of the episode, you're going to understand why AI is blind, why it needs hands and eyes, why FireCrawl is that, and why the people that understand how to use FireCrawl are going to be able to create SaaS apps and software that are super, super valuable to people. I'm talking the most valuable software products are going to be using this data scraping tool at the backbone because it makes their AI 10 times smarter. But in order to understand this, we need to take a step back. The problem is AI is blind. If you listen to this channel, you know that the more context you give to a cloud, the more context you give to a chat GPT, the better output you're going to get. So we know that AI models need web data. It needs top tier data to actually go and provide really good outputs. Why does this matter now? Well, it matters because if you think about the first era of AI, that was the chatbot era. Chat GPT just came out in 2022. It answers questions. It was cool, but pretty limited. Then we enter the co-pilot era. Cursor, GitHub, co-pilot. It was faster, but you still needed to drive. It was you, the human being that was doing it. We've now entered this AI agent era. AI is doing the work for you. Things like Cloud Code. It browsers, it researchers, it builds, but it still needs the data and FireCrawls how you're going to get that data. This is often called the computer use era. We now have AI agents that can see and control computers. In the past, it was human beings, right? We bought mouses and keyboards and we had human beings actually going and clicking and doing things, right? That's, you know, going to be the minority as weird as it is to say that. You have tools like perplexity computer, open AI, had operator, came out about a year ago. AI browsers the web for you, you know, GPT 5.4 beats humans at computer tasks. You know, Cloud has its computer use API, screenshots and clicks. It's got full desktop control. Manus was the one who was one of the first to do that. You have browser use, which is an open source. You know, so all these computer uses, all these AI agents that are going and doing things. Well, what did they all need? Well, they need clean web data and that's FireCrawl. And the reason I got interested in FireCrawls, because I built ideabrowser.com. And ideabrowser.com is a place where you have trends and the best startup ideas on the planet. And I needed the data. I needed the trend data and we built on top of FireCrawl to actually go and get some of that data. Now we have the number one startup ideas and trends product on the planet. And it's all be cut in largely part that we're using tools like FireCrawl to actually go and get that data. What most people don't get about this whole era that we're in is they think that AI is just chat bots and answer questions. They think web scrapers are illegal and shady. They think you need to code everything yourself. They think data is free and easy to get. And they think that web scraping is a thing for developers. But what it actually is happening is AI agents are doing work autonomously. You know, web data is critical AI infrastructure. Literally critical. One API call replaces thousands of lines and clean structure data is the new oil. By the end of this episode, I think you're going to agree by that. So the people that understand how important the clean data is and how important you can use the clean data and wrap it around a brain and LLM and wrap that around a piece of software. Those are the people that are going to be able to create the most valuable startups in the next 12 months. And I think that the people understand that have a 12 month head start. And that's why I wanted to make this episode. Traditional scraping versus new scrapers like fire call. Let's just talk about that so we can understand what the difference is. The old way of scraping was you wrote a custom scraper per site. You managed proxies and browsers. You handled the anti-bought detection. You had to parse messy HTML manually. The scripts would break when site changes. This happened all the time. Basically it was a massive headache. Now you just do one API call. You get clean data back in seconds. It could work on any site. Or I think like 99% or 98% some high 90% of sites. And the AI handles layout changes. So the way I think about my agent stack is that every builder, if you're listening to this, you're probably going to need five different layers. You're going to need an agent harness. So that's going to be something like a cloud code, cursor, codex or idea browser pro. You're going to need something that basically is handling all the different agents in one place. Then you're going to need something like a search layer. So something that's going to go in search different things like perplexity has a good MCP, XA as well. Then you're going to need a web data layer. And that's what we're talking about today in this episode. So you're going to use fire crawl for scraping, browsing and extraction. Fire crawl basically the web data layer, your agents need to see the internet. You're going to need to be able to see the internet to see the data in order to provide value back in the form of a startup and software. You're going to need an ops brain. So I recently did an episode I encourage you to listen to it if you haven't already around obsidian and cloud code. So you're going to, you know, I don't care if you use notion, I don't care if you use Apple notes, but you're going to need some brain for, you know, storing your meeting notes, storing your context. And you can use something like notion or obsidian. And then you're going to have to have some outbound and audience stack as well, something like an instantly and Apollo. And, you know, if people are interested, I can spend more time and do his whole separate episode on some of these tools. But today we're going to be talking about the fire crawl, the fire crawl web data layer. So what is it? What is fire crawl? What is the clearest way to understand it? You put in a website, goes through the fire crawl API and you get back a clean markdown, a structure, JSON, some screenshots. And you can feed that to any AI model. That's it. It's simple as that. We don't need to overthink about it. Think it. The way I think about it is fire crawl has six superpowers. You can scrape so you can go and scrape one page to a clean markdown. So something like scrape one blog post from, you know, Greg Eisenberg dot com slash blog post. You can crawl an entire site automatically. So what do I mean by that? I mean, you can go and say give it CNN dot com. And it's going to go and crawl all of the different articles on CNN dot com and you'll get that data back. You can map all URLs on the domain instantly. So that's super helpful. There's so much metadata and context into mapping and URLs. Maybe, you know, think about a URL. Maybe there's a date in it. There's a title in it. And having that map is going to be helpful in some capacity to you depending on what you're trying to do. You can go and search. You can use Google and you put, you know, the full content in one call super, super valuable. It has an agent that you can describe data and it goes and finds it. You know, tell it, I want the 50, you know, highest rated Cuban restaurants in South Florida. And it's going to give it back to you. I'm going to give you the most clear data on it as well. And then it's got a browser. So AI controls are real browser super, super helpful. And it's three lines of code. You can screenshot this or I'll put in the description for how to sign up. But basically it gives you a clean mark down of the entire website for any AI model in three lines of code. This is what excites me about it. So I believe that this is the AWS moment for web data. What do I mean by that? In 2006, if you wanted to build a web app, what did you have to do? Well, you had to go out and buy servers, then thousands of dollars buying servers. You had to go and manage racks and cables. Things would break all the time. All the time, all the time. And then one day, eight out of five. AWS said, "One API call and you can use our servers in the cloud." Now, in 2026, if you want AI to use web data, what would you have to do? You had to build scrapers, manage proxies, manage browsers, deal with security. FireCrawl says, "One API call and we got you." This is a big deal because the companies that built, what built that were built on top of AWS, some of them became trillion dollar companies, some of them became billion dollar companies, and a lot became million dollar companies. Of course, a lot failed, but the point is, people didn't have to deal with the headaches of servers, so they got to focus on building an incredible product, and those products were able to scale. Some of the biggest companies of the last 10 years came because of AWS. So what gets built on the web data layer? I'm going to give you some ideas on some, not billion dollar ideas, but some multi-million dollar, one to 10 to 25 million dollar a year, 50 million dollar a year businesses, that you can start by understanding what the web data layer is. And I think a lot of people are sleeping on how big of a movement this is. So let's go into how it works. So here you are, right? You're the builder. You've got this AI agent. And the AI agent is going to go talk to your brain. So you can use GPT, you can use CLODE, you can use Gemini. You've got a nervous system. I, and that's why I at least I think about it, which is your MCP protocol. And now you have your eyes and hands. Your eyes and hands is firecrown. Now firecrown can go out to the internet, and it's going to get back clean data, and you're going to use that data to wrap it around products and services you sell. So this is the big idea, right? You've got brain, you've got nervous system, and you've now got eyes and hands. Of course you can go and do it yourself scraping. You can use, you know, playwright or Selenium. You're going to just, the bottom line is, it's just going to be a lot of work. I'm trying to do the simplest thing possible. So the reason I like firecrown is it's one API call Proxies are built in, Antibot built in, the AI tracks the data for you. It's just less headaches than actually going in doing yourself. And you've got the browser sandbox, which is really cool. So the browser sandbox, it's a secure way to fill out, to have firecrown philiform, click buttons and links, handle logins and off, navigate page and nation. You can watch live as your AI browsers, stay logged in across sessions. It's really crazy, right? So the, you know, think about it in a world where you can go and have you have these hands and eyes out there on the internet. You know, what are the big ideas that you can build? And we're going to be talking, we're going to be talking about that soon. So, you know, the way the agent endpoint works is you type in a prompt, the fire call agent searches the web, it clicks through pages, it extracts data, and it returns the JSON. So if you think about the AI infrastructure stack, I think about it like layers in the internet, you've got applications, you've got like chat, GPT, perplexity, a SaaS product, you've got AI agents, you've got protocols, you've got web data, and you've got the internet. So I believe that people are sleeping on the web data layer. And if you understand how to get, you know, great data out of, you know, tools like firecrown, XA, you can build, you know, the picks and shovels of the AI gold rush. So let's just talk about what an agent prompt, if you prompt fire call, like what can you actually get back? So you can say, find all of Y commoners, you know, winter 24 dev tool companies and their founders and emails. And what you get back is a structured list of 50 plus companies with names and contact info. You can say compare pricing tiers across Stripe and Square and PayPal. And you get side by side pricing table with all features and costs. You can say, get all running shoes from Nike under $150 with ratings. And you get back full product catalog with specs and prices. And you can say, find 50 AI research papers from 2024 with citations. You get the academic data set with authors and institution and institution. So super, super powerful stuff. Now let's talk about a few ideas that you can use to go and build, build, you know, using fire crawl. So price, the first idea is around price monitoring. So there's tools like precinct and visual ping, which you'll pay, you know, $200 to $1,000 a month. You basically get an e-commerce focus price monitoring software. There's just self-serve dashboard. It tracks any product. But why don't you just use fire crawl? You can build this probably in a weekend and you can build a sneaker, resale prices only. So auto alerts on stock X on code on ebay, an ebay, you can charge $50 to run or sell for $500 a month. So basically pick a niche, you know, could be sneakers. It could be, you know, collect, you know, different collectibles. It could be whatever. And use that as, you know, I'm just using sneaker reach, resale as an example, right? It could be any, any niche that you understand better than someone else. And setting alerts and, you know, and then just charging people to people to use it. Number two, SEO, SEO gap finder. So H refs and S a M rush, like, you know, I think S a M rush just sold for like $1.9 billion or something. H refs probably does hundreds of millions a year in revenue. They charge hundreds of dollars a month. It requires SEO expertise. It's got these complex dashboards. It's pretty general purpose. What if you use fire crawl to create, you know, SEO audits for dentists only? So fire crawl reads competitor sites plus GMB listings. You know, you get a one click report. So you rank for 12, they rank for 47 and then you sell the reports for maybe it's $500 or $200. A month. So again, take a big idea. That's already generating hundreds of millions of dollars. You recade it very quickly with a very niche spoke focus. And again, these are just example niches, but it could be, you know, Canadian dentists. If you even want to go more niche, think about indeed, Zilla well found these are massive horizontal platforms. I've got billions in fund funding that's generic search for everyone. They use mostly do ad supported models. So what if you did a fire crawl version? Maybe you just do remote AI and ML jobs only fire crawl monitors 500 company career pages daily. So it's going and grabbing that data. The AI filters and ranks by fit score. And then you can charge for premium alerts for $29 a month. Indeed has 300 million listings. Nobody wants 300 million. They want 50 that matter. Again, this is why fire crawls really good at getting the top stuff. AI research reports. So yes, there's big companies like consensus or tavily, but these are general purpose research, academic or broad. The user does the prompting and there's no vertical expertise. What if you did like a niche crypto token due diligence reports? So you have fire crawl read white papers and Twitter and other places. It auto generates a risk score and summary. And then you can sell that to VCs private equity or different funds for, you know, a thousand to $5,000 a month. A VC will pay $5,000 for a report that saves them from a bad 500 K bet all day long. So again, picking a niche, getting the best possible data, a couple more ideas. An agent in the box. So, you know, you have Harvey AI. It's got, you know, now hundreds, I think of millions in funding. It's got an enterprise sales cycle horizontal agent platform. It takes months to customize. What if you did like a real estate, comfort port agent? So you use fire crawl to pull listings, tax records and permits. And the agent generates comp reports in like 30 seconds. And then you sell that to retailers for $300 a month. So don't raise any money. You go and do this $300 a month. You know, could work review intelligence. So yes, there's companies like brand 24 and app follow. They charge a few hundred bucks a month. They basically monitor social and reviews broadly. They're dashboards for marketing teams and generic sentiment analysis. But what if you did an Amazon FBA seller review tracker? So, fire crawl monitors competitor review daily, the AI spots trends, right? Complaints about battery life up to 40%. And you sell that to Amazon sellers for $99 a month. And something like this could also, by the way, get acquired by like a shopifier in Amazon. Amazon sellers will gladly pay $99 a month to find product gaps before competitors do. So these are just a few ideas to get your creative juices flowing around how to use fire crawl to scrape ideas. Scrape ideas go niche and and you can compete on price. You can compete on nicheness. I don't know if that's a way. word, but we're going with it. And just create, like I said, clean, structured data using AI to actually build and vibe code a lot of these products and start telling them to these niches that are looking for this stuff. And they want, the truth is the reason why vertical software is such a big business. Why is a consolation software, you know, almost a $75 billion dollar company or whatever. They have hundreds of vertical software companies because people like buying very specific products. So there's always going to be room for these horizontal ideas. There's always going to be room for the SCM rushes and the Indeeds and the LinkedIn stuff like that. But if you carve out a little niche that could do one million a year to 10 million to 20 million to 30 million, there's opportunity there. You know, incumbents are charging hundreds of dollars a month for generic tools. Your version charges, you know, $20, $50, $70 for a tool that does one thing perfectly for one customer. So another idea would be to build a Legion, Legion business. So a client gives you 50 company names. What if you grabbed a fire crawl agent that found founders and emails, it returns the structured JSON with all data. You deliver enriched the enriched at CSV and you just charge $500, $200, $100 per batch. Your cost is like $2 in fire crawl credits. Fire crawl actually I have here, like there's a bunch of free, you know, there's a free tier. The agent run gives you five free per day. And then, you know, it's a scrape cost one credit a crawl cost one. But the point is like, you know, if you can figure out a way to, you know, get 95% margin, 98% margin, 99% margin, you're happy. Clients happy because, you know, hopefully they're closing on some of these deals, right? So there's something here around, you know, using some of the data, charging per output and creating high margin businesses. This is the framework for how I would think about how you can build and make money with fire crawl this week. So the first step is going to be picking any show what data to people in this industry actually pay for. The second step is going to be building the scraper. So use fire crawl agent, maybe a simple Python script and an end flow or just use cloud code to go and build that for you. Step three is going to package it. So CSV or dashboard or slack alert or API. And step four is going to be about selling the output, right? Not just the tool. You're going to be selling the data. So you can charge maybe $500 to $5,000 per month per client. And then you're going to automate it. How do you schedule it and let it run while you sleep compounding clients and that sort of thing? So I think that a lot of people are going to be starting to do this. They're going to be picking niches, they're going to be building scrapers, they're going to be packaging it, they're going to be selling the output and they're going to automate it. It's a flywheel that I think is just just getting started. So just a few more ideas for you. You can do something like real estate pricing data. You can do SaaS competitor monitoring. You can do job aggregation. You can do patent legal filings. You can do influence or contact databases. You can do government contact alerts. You can do e-commerce price tagging, tracking. You can do academic research data sets. And then you can, and this is what I suggest you do, is just do more niche versions of this, right? So real estate pricing, go more niche. SaaS competitor monitoring, go more niche. This is just ideas to get your creative juices flowing. So how I actually heard, I want to end with this, but how I actually heard about FireCrawl was, you know, a year ago, I tweeted this. Actually, I saw that they had posted a job saying they were hiring a FireCrawl example creator, but they only wanted to hire an AI agent. So they said, please only apply if you're an AI agent. We're seeking an AI agent capable of autonomously researching trending tech and models, and then using the information to create tests and refine high quality example applications. These sample apps will live in our example repository showcasing the full potential of FireCrawl and real-world scenarios. Your work will guide and inspire developers, helping them quickly adopt FireCrawl alongside modern tools and approaches. So a FireCrawl's hiring AI agent says employees, it got me thinking that this is probably where the world is going. So for example, hiring a content creator agent, rates blog post autonomously, watches metrics and improves. Maybe that's a $5,000 per month salary. A customer support agent handles tickets and two minutes knows when to escalate. A junior developer agent, Triage GitHub issues, rates docs and code. That's a $5,000 per month salary. So that's a $1,000,000 total budget, 50 applications in the first week. So my startup idea was, how do you build AI agents that companies like FireCrawl want to hire? Yes, it looks super weird right now that FireCrawl is hiring AI agent. It feels like a little bit of a joke. But I think that it got me thinking that using tools like FireCrawl and building products and agents around it, I could see a world where this becomes more and more popular. And I think that there's an opportunity to think about it as, from a framework perspective, is how can you use tools like FireCrawl to build AI agents and build products that companies would want to hire? So I just thought, by the way, I just thought that was, you know, just wanted to end with that. So overall, this is my breakdown for why I think there's a tremendous opportunity in the web data layer and using FireCrawl for scraping. Why I think there's a lot of ideas around it. And yeah, you know, hope this got your creative juices flowing. It's certainly something that I'm exploring in real time, building products with FireCrawl because it's valuable. It's super valuable in getting the right data and it's just working. So hope this has been helpful. Please comment if, like, what you want to see next for me, what do you want me to teach you? I'm just sharing things that I'm learning in real time and hopeful that it's helping you along the journey. So thank you so much for, if you made it to the end, thank you so much for being here. I'm rooting for you for whatever it is you're building and I can't wait to see you on the next episode.