Code with These APIs to Get Rich in 2026 (full guide / examples)

What's going on guys? This is Oliver, formerly from Response AI and a few other software tools that I've since exited and now running roswell.dev, which is where I teach people how to code and make money with vibe coding and paperschedule.com, which is my own software that I'm running myself. In this video, I wanted to talk about the four best APIs to build a SaaS on, right, in 2026. Now, most AI tools this year are going to be making real money basically chaining together a couple of APIs, right? And they put a nice front end together and then they put a good price tag and that's it. So, what I'm going to be talking about in this video is how to make money by chaining those APIs. I've talked about this before and then basically picking the right APIs, wrapping it up in industries and then charging multiples and then this is the game. So, first of all, you're not reinventing the wheel, right? Other people are already spending millions doing this. So, you don't need to build a search engine, right? You don't need to build a web scraper or a transcription model or you know, social media integrations for some big app, whatever it may be. Companies have already done that and they will rent you the result of those integrations for a fraction of, you know, a cent per call, right? So, all of those things that they've done, all the data, all the servers they built, whatever it may be, you can just rent them. Your job is to recognize a boring problem, point the right API at it and sell the outcome to people who do not know or care about what an API is. Now, what makes an API worth building on, right? So, it does something genuinely hard. So, scraping through anti-bot defenses or transcribing audio really accurately, it's not some weekend project, guys, that we can build, right? As like a solo founders, it's something that these companies have spent millions doing and then obviously they're making money by selling it back to us as an API. The difficulty is your moat against the customer ever doing it themselves, right? So, it has a free or cheap entry point as well. You should be able to build the whole thing as like a hobby project and then start charging as you get customers. And the pricing is ideally per call. So, your customer pays for an outcome. That asymmetry is where the margin lives, right? So, they pay for the report or the transcription or the integration or the data, whatever it may be. Um and you make that middle ground. So, you're paying an API $10 and you sell it to your customer for 20. You can wrap it in context that makes it sticky. So, just a random API is just a commodity. There's plenty of them. But, the same output filtered through their business rules or your business rules is something they will not cancel. So, let's consider a transcription app that um is for roofers and contractors that transcribes meetings you have with customers, right? Um assuming that obviously you have that know-how and you know what information you're looking for when talking to a customer if they're trying to sell them a roof, that's the difference between being a general transcribing tool and being, you know, RoofScribe, which is this new, you know, I'm making it up, but a new app that helps roofers and contractors transcribe the meetings they have with clients on site. So, you see how those things work. You have to have some specific knowledge and then apply it to those um industries. So, part one, right? So, let's talk about the first API that I have built a lot of different tools on for myself or for clients. Exa is a web search API, but not the kind that returns just like random blue links. It does semantic search, which basically means that it understands the meaning of like a web page rather than just matching keywords. So, its standout feature is um find similar websites, right? So, hand it a URL like apple.com and it returns pages and websites that are conceptually like it. And it also holds an index of billions of people profiles and around 70 million companies, right? Plus dedicated search modes for news, finance, code, whatever it may be. Now, the margin you're working with, so there's a free tier of 1,000 searches per month. So, that's plenty to build a product on or validate a product or just build a demo for clients. For paid, it's about $5. So, in my experience, it does work out at about $5 per 1,000 standard searches. And then obviously, it's more expensive if you pull full text of a page, it costs about a dollar per 1,000 pages. You see how much you can scrape for that cheap. The bottom line is about half a cent to a cent for a search. Then the content and that is your cost. So, idea one for building on this API is look-alike lead lists. So, for B2B agencies that want more clients like their best one. So, what it does, it takes an agency's favorite existing client, it feeds that company's website into the find similar feature of Exa, returns 50 look-alike companies, and then it enriches each enriches each, sorry, with people profile data to find the decision maker, right? Now, who's going to buy this? So, agencies, outbound teams, founders, solo founders, startups, anyone who has ever asked, "Where do I find more clients like my favorite one?" And the math is this, right? The cost per list is 10 to 30 cents in API calls. The price is $99 to $199 for a set number of lists, whatever it may be. The pitch is that you can find more of your favorite clients on autopilot. And then this is without even considering the fact that you could plug it into a B2B lead source like Apollo or like People Data Labs or something like that that allows you to actually get the data of those websites. So, imagine if I said to you, "XYZ agency is your favorite kind of client. Here's 100 more and here's their mobile numbers and emails." That's super valuable and people buy it all the time. Industry intelligence digests, right? So, what it does is this is idea two, it picks a narrow audience like healthcare consultants or commercial property investors, right? So, let's assume that we're selling to property investors. You use X's news and answer modes to pull and summarize everything relevant to that niche every single morning and send an email to them. So, who buys it? Busy professionals who currently scan 30 sources every day like Wall Street Journal or real estate, you know, real estate magazines, whatever it may be. And instead of doing that, you can just feel informed every morning for a set price. And it costs cents to assemble and then you can charge something like this like a newsletter I would charge at $29 even like $19 or $15 because the the narrower the niche, the higher the price you can pay. But the pitch is effectively, look, one email a day and you never miss anything in your field again. But it's not the same as having like just be just, you know, reading the news, it's the fact that you have niche-specific digests every week or every day even. for something like the content side and SEO side, idea three would be SEO research and content, right? So, what it does, use find similar, uh so the find similar feature on a target article to surface the strongest pieces in a space, summarize the angles, gaps, structure, then feed it to something like Claude or ChatGPT's API to write the articles, right? So, freelance writers will buy this, more marketing teams will buy this, content agencies, anyone writing briefs for someone else to execute or blog posts, etc. The math is that it costs cents to report something. You sell a per report or as a flat $49 a month for X amount of reports. The pitch is two hours of manual digging before every brief or before writing and all of that's automated. Now, let's go into part two, which is the second API that I would build on and have built on, especially for paper schedule, my app, I've built with Firecrawl and Exa. So, I've literally built apps that make money with these two so far, right? So, Firecrawl scrapes [snorts] and crawls websites and hands you back clean structured data, right? So, you can put it in an AI model like ChatGPT or Claude or Gemini or you can put it in a spreadsheet, whatever it may be. It handles all the really shoddy, nasty stuff about scraping the internet. So, you've got JavaScript rendering, proxies, anti-bot defenses, lazy loaded content, all these things. In four moves, you scrape a single page, you or you can crawl a whole site, map all the URLs in a domain, and then extract whatever you need. Now, the free tier is 500 credits per month. Then, the hobby plan is like $16 or so for 3,000 credits, 5,000 credits, whatever it may be. And then, the standard is around $83 a month or yearly, it's like £62, whatever it may be. The real numbers a simple scrape is well under a 10th of a cent, and a complex extraction runs about half a cent, right? Open source, you can self-host this once you scale and cut the cost. But, scraping the big social platforms is restricted, so I would use something like Apify to scrape like, you know, social media. So, idea one using Firecrawl, e-commerce operators who currently check their competitors manually and their prices manually, right? So, what it does, it crawls a competitor's products or pricing pages daily, detects what changed, and then alerts the owner the moment a price drops or a new product appears. There are literally tools, guys, making 10, 15 to 100,000 pounds or dollars per month doing this because you are crawling a competitor's product, and then you're effectively eliminating hours of research every week from those customers, right? Now, who buys it? Online retailers, D2C brands, marketplace sellers, anyone whose margin moves when a competitor moves, okay? Now, the maths is a few cents per competitor per month and the price is $49 to $149 a month, right? Depending on how many competitors they track. Typically, what I found is that um these daily sort of like competitor pulse apps, um they might charge $49 or so and you can you can effectively track say 10 competitors, [snorts] But things are getting cheaper and cheaper to do, so you just have a think. The pitch is the boring task they dread is made invisible and automated. That's as simple as that. The next one is inspired by a multi-million dollar business called Chatbase, which I really, really like. Now, idea two based on Chatbase is the chat with your website for local businesses. Basically, you can use Firecrawl to crawl a small business's entire site and then turn it into text and then feed it to a language model. The result is a support chatbot you can put on their website and it's trained on the business's actual content and questions and FAQs and customer queries, etc. A customer customer can ask the chiropractor's website, "Are you open on Saturday?" and get the right answer. Or have um you know, "Do you support uh pets on the premises?" Whatever it may be. Who buys it? Local service businesses, chiropractors, dentists, gyms, restaurants, salons, anything. Anyone whose customers ask the same 15 questions every single week or ask questions that are kind of left of field. Like say a tanning salon, you might say uh you know, you might ask a question like, "Do you do um couple deals? Or do you do discounts for three people or more?" These are more complex questions that usually require a human to answer, but you can save time by allowing them to do that without, you know, a human intervening. The maths is that you crawl the whole site cuz like 20 cents plus tiny ongoing AI costs, zero um costs for keeping the data places, and then you charge them $39 a month to actually have the chatbot on the website. This is not like made up, guys. These are real businesses, and if you look up Chatbase, they make an absolute killing. They literally make millions of dollars a year doing this, and it's all based off of the chat with your website kind of um approach, basically. Now, the pitch is, I want you to stop answering the same five questions every single day, and just automate these customer support aspect of your business. Now, idea three is the vertical data extraction sold as a feed. So, you pick a scattered industry, and you sell the tidy version. So, what it does, you can use Firecrawl to pick an industry where information is scattered across many websites, right? So, you've got job boards, real estate listings, supplier directories, whatever it may be. Um like a hiring uh or like freelance websites, those kinds of things. Use Firecrawl's extract feature to pull the same structured fields from every site into one clean data set, then you sell access to that data set, right? So, one thing I did this for was a previous startup I owned called Capstone, and it was a list, excuse me, of thousands of investors, literally tens of thousands of investors who invested in series A, series B, series C, angel funds, seed rounds, whatever it may be. And the companies that bought it were people who wanted to find investors and their contact data. This is a similar thing. I just extracted data at scale. The companies in these verticals who need the data but can't collect it themselves, like brokers, analysts, sales teams, whatever it may be. Now, the math here is that discipline matters. At half a cent per page, a 1,000-page report might cost $5, right? You sell it for $50. Always know your cost per outcome, guys, when you're building these API-based tools, okay? And the pitch is the data set they need but cannot build themselves and cannot find. Part three is Deepgram, okay? So, I've built on Deepgram, there's thousands of multi-million dollar companies that have used Deepgram or similar APIs to build stuff. And this is probably the most potential money-making one for me in in my opinion, right? Deepgram is an audio layer, right? In so many words, speech-to-text models that transcribe recordings and live audio really, really quickly. And you can build basically anything you want on it. So, there's text-to-speech, audio intelligence, voice agents, whatever it may be. The audio is cheap to process and easy to charge a premium on. So, there's a free tier of $200 credit, then it's pay-as-you-go. And it's something like 46,000 minutes of pre-recorded transcription. You could run a small product for months without paying anything, guys. Then, text-to-speech is like $30 per million characters. The bottom line is that it's cheap, and the voice output is the premium, right? So, idea one is niche meeting notes and call summaries. So, do not build transcription for everyone. We've got Granola, we've got Fathom, um we've got Otter. All of these different tools already transcribe meetings. Pick one type of business instead. Like sales call notes that auto-fill a CRM, right? Therapy session summaries formatted to a clinician's template. Site visit notes for contractors who talk into their phone or talk to clients between jobs. Who buys it? Whichever vertical you pick, guys. The vertical formatting is the moat, okay? It's a transcription for everyone is a crowded race to the bottom. Everyone's doing it, and some people are even offering it for for free, you know? The math is that an hour of audio might cost 20 cents to transcribe, and the price is $30 to $99 per month per user, whatever it may be. The pitch is that the hour of typing after every meeting is gone. And the typing during a meeting. The next one is podcast and video repurposing. So, for Deepgram, you can transcribe the episode of a podcast with Deepgram, use a language model to spin out the notes, timestamps, quote clips, social posts, whatever it may be. And basically, you take someone who's a podcaster and you output a full week of content for them with one recording. Who buys it? Podcasters, video creators, anyone with a recurring long-form show who hates the pro post-production grind of formatting stuff for their social media. The math is that a 2-hour episode costs $100 to transcribe and process, and then you can charge $49 to $99 per month with X amount of, you know, processes. The pitch is that you turn one recording that you've already done into a week of content without lifting a finger. Someone like me literally needs this right now. I need to find a way to repurpose this content that I do all the time on YouTube and put it elsewhere. So, that's where I'm literally could be, you know, I'm the first customer effectively. Idea three is AI phone answering for trades and clinics. So, again, this combines speech-to-text, text-to-speech, and a language model through the voice agent API you can use on Deepgram, and it builds a phone assistant that answers calls, books appointments, takes messages for, say, dentists, plumbers, salons, physios, chiropractors, um you know, massage therapists, whatever. Any business that misses calls when staff are busy and loses bookings and money to voicemails. The math is that about $4 or so per hour of calls, even heavy usage is a few dollars a day per and the price is $200 to $500 per month cuz you literally you are literally replacing a employee. And I know that there are companies that are doing this for cheaper, like, you know, $79 to $99. So, consider the market research first. The pitch is that you never miss a booking again. One recovered appointment for, you know, Invisalign or for braces might be worth $5,000 to a client. And and you've, you know, you've you've saved that for them for, you know, $100 a month. Part four, so in the sort of latter stages now of this of this of this video, is Post EE's and Zeeno. So, basically, Paper Schedule, which is my current app, you literally I literally use Zeeno to run the entire app, okay? So, I this is this is me using one of these APIs to literally make money, okay? And basically, Zeeno and Post EE's, you can publish to every social platform in seconds, okay? They both solve the brutal problem of posting to X, Instagram, LinkedIn, TikTok, YouTube, whatever it means. Instead of dealing with 15 different APIs and authentication systems and sets of rules from all these different platforms that break constantly, the integration with Post EE's or Zeeno allows you to do it once and with one simple interface. And Paper Schedule, my actual app that makes real money, all you do is log in and it says, "Post to a dozen platforms at once." This is you just Zeeno. Like, steal my business, guys. Like, you can you can literally take that now. You know that I use Zeeno to create a social media scheduler and then you can go and do it. And the social API is just the pipe. Your customer never sees Zeeno. It never knows that you're using it. The social API is just the pipe, right? Now, I use Zeeno because it's like a per seat pricing nowadays. Like, first two accounts are free and I think like it's priced per account per month, whatever And Zeeno pricing is free for those first two and then usage-based from $6 per account per month, whatever it may be. I think it's even less. I got it for like $3, right? Post EE's is open source, but the cloud version is from around $29 a month and you can do auto posting, connects to all these different tools like Make and N and I would pick Zeeno because there's no servers and it's easier. It's like not as it's not as challenging for me, anyway, as opposed to these. But, Postie's is insane and it's growing at a really crazy rate and there's zero ongoing platform costs and control for the data, right? So, let's talk about what we'd build on this, except what I've literally built. So, Paper Schedule is literally a social autopilot, right? But, you can pick the niche, ask three questions, and then forget social forever. So, what it does, pick a real estate agent, a restaurant, a gym, ask the owner three questions about their business, generate a month of platform native posts with the language model, then schedule them all through Zeeno or Postie's. who buys it? Owner-operators of small local businesses who don't want to do posting. They want to stop worrying about Instagram, whatever it may be. I know that I can charge about $49 a month for this with Paper Schedule and you can charge the same or you can charge more. The pitch is that you stop worrying about posting on social media forever. Idea two is the same, but a white label scheduler for agencies. So, you can build a clean branded dashboard on top of Zeeno or Postie's. Then, you say to an agency, "Would you like to manage every client from one place with approval workflows, per client analytics?" And who buys it? Social media agencies or social media teams. You charge $50 per managed client and then the underlying costs are fraction of that, guys. The pitch is that one dashboard for every client will help you succeed. Moving elsewhere becomes a nightmare that they a nightmare cuz they don't want to have to move from your platform. Idea three is content repurposing that publishes itself. So, Postie's talks about this a lot. The guy who founded it, he's great. Um, I've forgotten his name now. I'm so sorry, but Postie's, p o s t i z, and Zeeno, both of those founders are great and talk about this a lot. Take a blog post, turn it into platform native posts or LinkedIn posts or videos, and then auto publish on those schedules. So, customer connects their blog or channel once. Every new piece of content automatically becomes a week of social posts. I do this with my YouTube. So, if I've posted this YouTube video now, I'm going to repurpose it with this flow to do like 100 tweets, whatever it may be. And the math is that there's penny per pennies per piece of content. Charge like $79 and then distribution on autopilot the posting API makes automatically actually happen, right? the real money is in chaining, guys, okay? So, every link you add makes it harder to copy. Every idea so far uses one core API. But, you're not selling APIs, you're selling outcomes. The customer judges the price against the hours you've saved them. So, two examples of chaining these APIs together would be the thought leadership engine. So, we're using Exa plus um Claude plus Zeno. So, Exa monitors news and discussions in the customer's industry every day. Then the language model like Claude or Gemini or ChatGPT analyzes what matters and drafts takes on it every day, then publishes it to Zeno. Paper schedule, my app literally does this already. So, I I I'm showing you exactly how this works. It's literally just Exa, an LLM, I use Gemini, and then Zeno to post it. Now, who buys it? Well, all of my customers, busy founders, executives, whatever it may be, startup founders. And the math is that you could charge way more than I do. You could charge up to $200 a month because it's literally like having a full-time social media advisor. Excuse me. And you look active and informed online and you make loads of money from your content, right? Chain two [snorts] is the creator pipeline. So, Deep Gram plus LLM plus Post E's. So, Deep Gram transcribes your podcast or your video. A language model writes the notes and social copy. Post E's schedules it and publishes the whole batch. Podcasters and video creators who hate post-production and like distributing their content. And at $2 per episode for you to manage with API costs and the pitches look one recording in a week or a month or a year of distributed content out. So, how to start this week, guys? Stop researching, pick one. Pick one API and one boring industry you already understand. So, I understand sales, I understand marketing, I understand copywriting, I understand YouTube, right? Make it work by hand first. So, run the core pipeline once before you build any front end. So, in other words, make sure that you can take a podcast, transcribe it, then turn it into content in one go. Get five paying customers at $29. A customer who pays is the only validation that counts, right? Do the margin math before you price because I ran into this issue. I started pricing Paper Schedule really cheaply and then I started losing money because the APIs were costing me too much. Cost per outcome, not per month. If you cannot work it out, you are not ready to charge money for it, okay? I would always try and plan it to be like credit-based or like a set amount, but just make sure that it never costs you that much. So, now, if I charge $49 per month for Paper, I know that it costs me like $10 to actually manage a client, okay? Now, the cost per outcome, not per month. Own the glue, business rules, and niche formatting are the reason they stay subscribed like the roofer who only wants to transcribe his meetings. APIs can't offer that cuz they're so generalist. The big lesson, you are a connector, right? Nobody cares how many APIs you're using under the hood of what it costs you. No one cares about how the car engine works as long as it drives. Your job is to see a tedious job, then create a solution for it, right? A specific type of business hates doing that job. You find the APIs that handle a slice of it each and you stitch them into a pipeline that produces a finished result. And the next profitable SaaS that you build or I build will be built by one person who picked a dull problem, pointed three APIs at it like LEGO pieces, and then they built it, and then they started charging for it. If you guys want the full playbooks for managing this, running this, making money with software, go to roswell.dev and I've written an entire book. It's like a hundred odd pages with dozens of things, prompts, scripts, starter templates, code, everything. Um and ask me if you have any questions in the comments, but take care.