Transcription
Hey guys, in today's video, we're going to break down the question that's on every developer's mind in 2025, which is what coding tool should I use, Claude Code or Cursor?
Well, in the rest of this video, I'm going to break down exactly which tool you should use and why, so that you can save as much money as possible while still cranking out code with AI at lightning speeds. And this is going to help you whether you're coding on a budget or balling out spending hundreds of dollars a month on these coding tools.
And just in case you're curious, why the heck should I listen to this guy? Well, my name is Brandon Hancock and I run a school community with over 10,000 AI developers. Plus, I run shipkit.ai with hundreds of developers where I teach them how to build real-world AI applications in days instead of months. And I've spent thousands of my own dollars playing with these tools. So, I have seen every situation. And in the rest of this video, I'm going to help you pick the right coding stack for your own situation.
So, here's a breakdown of the three different topics we're going to be covering in today's video. First, I'm going to be sharing with you guys my preferred AI coding tech stack so you can get a feel for what's possible, and I'll briefly explain why I'm doing it.
In part two, we're going to go deep into actually understanding the rate limits, usage, and cost per request with Cursor versus Claude Code because they've made it as confusing as possible to pick. So, we're going to clear it up so you can fully understand exactly what you're getting bang for buck when using these different platforms.
And finally, in part three, I'm going to be sharing with you guys four different profiles to help you pick which tech stack you should use based on your current budget and needs. So, we're going to go all the way from you're coding on a budget to you're an absolute baller happy to spend hundreds of dollars per month. So, we're going to break it down into four different profiles so you can pick the best one for you.
So, that out of the way, let's go ahead and hop into the video. Oh, and because you guys are awesome, I've put together a Claude Code cheat sheet that you can grab down below in the description. And in this cheat sheet, I put together how you guys can set up Claude Code, how to use all the proper shortcuts, and most importantly, how I configure Claude Code for me and all of my developers inside of Shipkit, where we teach them how to actually make Claude Code feel just as easy to use as Cursor. This thing's absolutely awesome, and I know you guys will love it. Just check out the link in the description below. But enough of that, back to the video.
All right, guys. So, let's dive on into my current AI coding stack, which might be a shocker to you guys, but I'm actually doing both. And in this section, I want to explain why I'm doing it. And then in part two, whenever we actually dive into comparing all the different plans, you're going to have a much better understanding. So, I definitely recommend to wait around to part two so you can see some of the things that are happening behind the scenes and why I'm picking two so the math makes sense.
Okay. So, quick bit of background. At this point, I am using both Cursor and Claude Code to crank out AI tokens like there's no tomorrow all day long. So, specifically, I'm using Cursor on the $20 a month plan and I'm using Claude Code on the $100 plan. Now, here is why I have both.
Okay, so Claude Code refreshes on a 5-hour window. And for me, I'm literally here at this desk. I probably should go outside more and touch grass more often, but I'm here cranking out YouTube videos, working on own startups, multiple. I'm helping out customer projects. I'm literally working on documentation. I'm doing everything. And all of my work has to be done with AI cuz I want to get more done faster. So, because of that, I'm using Cloud Code on the $100 a month plan. The reason why is I get 200, like it's the math basically breaks out to around 225 requests for every five hours. And you know if I'm starting work at 9, well, that gives me a morning session, an afternoon session, and if I want to push it, I can even do an evening session. So in a day, I'm getting 600-plus requests in a day, which is insane. I am never really hitting this usage limit even with heavy use. So that's why I'm doing this. On a good week, I will roughly hit around 60 to 80% of this limit. So, I'm never going over with heavy usage throughout the entire day. So, absolutely love using Claude Code.
Now, at the same time, you might be wondering, okay, well then, like, if you're not hitting the limit on the $100 plan, why are you also spending $20? Like, why would I even need to worry about this? Phenomenal question. The main reason why is because Cursor gives me the ability to do a few things that I can't inside of Cloud Code. Specifically, Cursor allows you to use the 1 million context window when working on bigger tasks. So, specifically, there are sometimes when I'm doing AI development and I'm trying to work on a monster task to where I'm trying to like, hey, you need to understand the entire backend. Heck, I have this entire other reference project that explains exactly how I'm trying to work on this specific task. So, I just have a monster amount of context that I want to put into one window to help implement something in my current project. Well, I can't do that right now inside of Claude Code. Claude Code has a max of 200,000 tokens by default. So, there's just no way for me to get access to that other 800,000 tokens. So, whenever I'm working on a monster task, I need somehow to achieve that. And the way I'm doing that right now is by using Cursor. Specifically, it's kind of a pay-to-play game, but in Cursor, you have the ability to get that maximum context token windows like, but you have to pay $20 a month to start even be able to have access to these features, which is crazy because even though you're paying the $20 bucks to actually get access to maximum content windows, what's wild is once you do get access to, you know, 1 million tokens, the part that's crazy is they then you have to use max mode. So, you're specifically using max mode and you're paying the $20 per month, but on top of that, what's wild is you're also then going to pay for API tokens on top. So, at this point, I'm basically using Claude, um, I'm using Cursor as a fallback for whenever I need to work on a monster task.
The other reason why I'm using Cursor is because I have the ability to do unlimited tab completion. So if I ever am actually like coding, which is very rarely, I'm usually using Claude Code or Cursor to actually do all the tokens and actually do all the AI work for me. But sometimes if I'm actually in a file and I'm like, "Oh, I know exactly what I need to do here." Usually I can like code a few lines and then boom, it helps me finish the rest.
And then finally, the other reason why I like to use Cursor, this is strictly more for an experimental standpoint, but I want to have access and play with all the models. So I already have to have an IDE. I already have to go ahead and I want to use maximum context windows. So, it's nice that Cursor also allows me to chat with everything from the new GPT models when they come out, Gemini models, like I get to try everything. So, Cursor is more of my experimental and big boy complex task. And outside of that, I'm using Claude Code religiously for everything.
With that being said, I do want to quickly mention how I am using Claude Code because I don't like to use Claude Code the way that most people do, which is they're, you know, pretty much just running Claude Code down here in the bottom of their screen. I really dislike this. What I like to do when using Claude Code to get the most out of it and the best experience, what I like to do is I have set up a bunch of shortcut keys for my Claude Code instance. So it feels just like I'm inside of Cursor where I can over here on one side I can have all of my actual projects set up on the right. So I have some code in one window, I have AI chat in the other. And it's so nice cuz I can open up as many chat windows as I want. I can work on many AI tasks as I want or in each AI task I have my own Claude window helping me work on a specific task. And I also have some other shortcuts to where I'm renaming tasks. So like this is going to be my setup task. Then over here, this is going to be my like landing page window. So this is where in each one of these I'm working on a separate task using Claude Code in its own conversation. So this is how I like to use Claude Code to get the max out of it. And don't worry, everything that I've just showed I actually in the cheat sheet that I have downloaded for you guys in the description, you can see exactly how to set up your own environment to have all these same shortcuts. So when you're inside of Cursor using Claude Code, it feels more natural. It feels like just how you normally would whenever you have the sidebar and you're actually working with agents and having agents do all the work in your actual files. So, it's that same kind of feel, but we're doing it now with actually just Claude Code because we've set up some customizations to make it just feel natural. Don't worry, download it for free. I've been using it. People inside of Shipkit have been using it and they've absolutely been loving it and it is really a step above normal Claude Code once you get it working in your IDE.
So, with that out of the way, what are we doing? Well, at this point, now that you've kind of seen how I'm using Claude Code, I want to break down in part two. I want to show you guys some of the core complexities of like, you know, like we're going to go deep into some math and to actually break down how each one of these tools calculates tokens, calculates usage, so that you can truly understand when you're putting, you know, your hard-earned money into these different systems, how much output you're actually getting, because you're going to be shocked to find out some of the ways that Cursor charges you versus Claude Code. It blew my mind once I did the math and I hope it blows your mind, too. So, let's go ahead and hop over to part two.
All right, so let's dive into the numbers so we can understand how the heck do all of these AI coding tools actually charge you money and what do you get out of them. So, in this one, we're going to compare Cursor, Claude Code, and I've thrown in a little extra bonus for WinSurf. Then once we understand how they actually work and what limitations do you have on the different plans, then I want to show you some of the math I've broken down so that you can see what you can expect to get charged per request so we can figure out what is the best AI coding tool that we should be using. So let's dive on in.
So we'll start with Cursor because it is the easiest to understand. I will say Cursor has done a phenomenal job of clearly articulating. Here's what you're going to get for this amount of money. And let's actually break it down how it works. So, when you're in Cursor, they are going to charge you $20 per month. And what that's going to give you is 500 prompts or requests where it doesn't matter if the message you send in is, "Hey, what's up?" Or you send in a monster file with all sorts of custom instructions with images and like everything. It does not matter. You get charged the exact same amount of money. and you get 500 of those requests per month. That's what you get on the $20 a month plan.
Next, what they have is the $60 a month plan and it gives you everything that you get in the Pro plan, which one of the big benefits was maximum context window, the 1 million for models like Claude 4.5, Sonnet, ChaBT5. So, it you get models, all those models at the maximum context window, at least the ability to use them. But on the $60 a month plan, what they're going to do is charge you or give you access to three times the usage on all those models. So in the past you had 500 requests, well now you have 1,500 requests.
Now what is interesting though is what you'll start to note it is we're starting to form a linear relationship where the more money we put in, it's just we're getting, you know, we're not getting any benefits for bulk buying. So, normally what you would like to see is instead of like I put in $1 or $100, what you would expect to happen is actually to see some sort of like bulk discount to where every additional token I buy, I get it more of them. So, if I give you $100, I would expect to get a discount because I've given you more money. But that is not how it works with Cursor on the $20 a month plan or the $60 month plan. It is just a very straight line. More dollars in, well, you linearly get more money.
Now, what they do have is the ability to buy the Ultra plan, which doesn't give you just, you know, a little bit more extra. It gives you 20x extra, which is crazy. So, if you know, going back to our chart, it's no longer like this to where at $20 we get this, at $60 we get this, and at $200 we get this. It's more like this to where it's just, you know, it's some hockey stick growth, which is what we would like to see. But for $200, that's kind of a, you know, that's a steep price to pay. But you basically end up getting at the end of the day much cheaper volumes in amount of requests once you go up to this amount.
Okay, so that is how Cursor works. Very straightforward. You get a fixed number of requests whether this big or this big. Key thing to remember.
So now let's dive into Claude Code because this confused the hell out of me the first time I was watching it and looking at it because I was like, wait, what am I getting? How much am I going to get? Like, none of it made sense. So let's actually break it down. Claude Code. The key thing that you have to remember is you are getting basically limited on multiple things. You're getting limited on a 5-hour window and you're getting limited on a total number of weekly usage. So those are the two things that you have to keep in mind today for the, you know, the current 5-hour slot. So the second I send my first message, boom, the timer goes off and the first 5-hour window starts and there's only so many requests I can send in that 5 hours. Then on the broader scheme of things, there's a weekly usage. So at the beginning of every week, you know, the timer starts and you have a limit. So it's a little confusing on what you can and can't do with Claude Code. And it's very confusing in that 5-hour window how things are actually tracked. My understanding is you are getting charged per runtime. So Claude Code's thinking about your stuff, you're getting charged. So it doesn't matter if it cranks out, you know, 20 tokens or it cranks out a thousand, you know, in 10 minutes, you're getting charged on those 10 minutes of runtime from my understanding.
Now, here's the math and the mental gymnastics that we have to do as Claude Code users. We have to figure out on average, well, how many requests do I get in a 5-hour window? So, what that kind of breaks down to is we are going to get roughly 45 prompts per 5-hour window. So, that's what you get on the Pro plan, which ignore the $17. That's the annual plan. It really costs $20 per month to get access to this right here. Okay.
Now, the next step up from the Pro plan, which is going to be the Max plan. Now, what's interesting is there's actually two separate Max plans. There is a $100 Max plan and then there's a $200 Max plan. They give you the same thing except one of them gives you a higher amount of usage. So, let's actually break this down. So, on the Max plan, when using Claude Code, you get, you know, instead of spending $20, we're now spending $100. So, we're paying 5x and we're getting 5x. And what's also nice is we're not just getting 5x on, you know, the amount of tokens. Like, all of our limits go up. So what I mean by that is in the past instead of us having a limit of let's just say the 45, you know, requests per 5 hours, we now have 225 requests per 5 hours. And then if we were just to extrapolate it's very hidden what their weekly usage limit is. Like it's very hard to figure that out. But if let's just say it was 5,000 on the Pro plan. Well now let's just say it is 25,000. It's 5x. Everything is 5x. So that part's I can, I can do that math. I can multiply by five. But the hidden numbers underneath are a little bit confusing. So this is what you get on the Max plan.
And then finally on the Max plan, on the 20x Max plan, what's nice is it's starting to form that nonlinear pattern. So just to be very clear what I'm trying to say here. So at the, you know, at the $20 a month plan, we get this amount of usage. On the $100 a month plan, you know, we get 5x the usage. But what's nice is the second we go to the $200 a month plan, we're actually way up here where we're getting way more output. So we're getting 20x the output of what we were getting over here, but we're only spending 10 times more. So just the math, we're spending 10 times more, but we're getting like almost what is it? 20x. So spend 10 times more, but get 20x the output. So, that's where we're starting to get the bulk discount. And you could expect in a, you know, over here on the Max plan, if you're spending $100, you would expect roughly 225 prompts per 5-hour window. Or if you're on the $200 a month plan, you could expect to get around 900 prompts per hour, which is an insane amount. Like, you're going to see this when we start to break things down here in just a second in the math. Like, you have to be an absolute machine and pretty much automating everything about your life to actually start to hit some of those $200 a month limits. Like, it's insane what you would have to do to hit some of those numbers.
So outside of that, I do just want to quickly cover WinSurf because I think this is an alternative that some people like and they pretty much have one tier, which is $15 a month per user. And when you use that, you're getting 500 prompts per month. How does that compare? Well, going back over here, it's $5 cheaper than Cursor to get, you know, basically the same thing. I will say I've noticed Cursor, at least when I have used WinSurf a while back, I felt like the Cursor agents under the hood just did better than WinSurf, personal experience, and that was from multiple months ago. So maybe WinSurf has caught up, but I was happy at the time to pay $5 more to have what I felt like was a better agent experience.
Okay, so now that we have covered at a high level what the different platforms of these AI coding tools, what they charge, how they work under the hood, what I want to do now is actually let's dive into some of the math so that we can figure out and like, well hey, based on am I a light user, a heavy user, what am I actually paying per request so that we can be on the same page and start to pick which one of these tools should we use based on our situation. So, um, we're going to go through the math and it's going to be, you know, I've broken it all down, so it's super easy to understand, but I think you're going to be pretty shocked with some of the results.
All right, guys. So, let's go ahead and dive into analyzing all of the different AI coding tools side by side so that you can pick the one that is best for you. And in this video, you're going to see we're going to start with the most expensive tools and work our way up to the most affordable. So you can see more as we get more affordable per request, we actually start to pay some of those higher prices. And what I really want to do is break down some of the math behind picking out these different models because I've, I've said this whole video, we're comparing apples to oranges because Cursor and all the other tools are per request. Claude Code is on a time basis. So it's like, do you get 100 requests, 200 requests? No one really knows. So I'm going to show you some of my research and how I got the numbers I did to help you guys pick the best model. So, we're going to speed through this so you guys can see everything in action.
So, we're going to start with Cursor. Cursor is by far the most expensive tool to use. It doesn't matter if you're on the Pro plan, Pro Plus plan, you are always paying 40 cents per request when using Cursor. Now, what's interesting is it's actually not always 4 cents per request. Sometimes when you're using thinking models like Claude 4.5 Sonnet, they charge you twice the amount of tokens. Like, it counts as two requests basically. So, you're actually not paying 4 cents, you're paying 8 cents when using Claude 4.5 thinking, which is crazy. That means for every time you press enter on your keyboard using that model, whether you say hello or you put in a monster prompt, you're paying almost a dime to use Cursor, which just blew my mind the second I actually saw the math behind it, which is crazy. So, one of the things though is yes, it is the most expensive one to use. However, it is the most convenient, beginner-friendly, all-around, like ease of use is insane and it has some amazing features that aren't offered anywhere else. So, when using Cursor, you're going to have the beautiful IDE that's going to let you pick and choose, view all the edits, view all the changes, undo them, accept them, decline them. You're getting that functionality. You have the ability to hop between all of the latest and greatest models. Claude, Gemini, GPT, like they have done a phenomenal job on the product. However, they're charging you for it. So, that is something that you do need to know when using Cursor. They are charging you for all that amazing convenience that they're providing to you.
Now, let's keep going up the list. So, we next up have Copilot. I just wanted to show this because I know there's a few developers in my community who have Copilot. So, I just wanted to show everybody what that actually looks like on a request basis. So, when you're using Copilot, they have the most affordable plan of $10 per month. However, you're not getting the best bang for buck out of that $10. As you can see, you get 300 requests in a month, which is honestly, like, I feel like I've done that in a day sometimes. Like when I'm just in an all-out coding session, I swear I could hit 300 fairly easy in a day. So, I never have used Copilot, but I know some people have. So, I just wanted to let you guys see where that stacks up compared to everything else. And maybe you want to switch.
Now, what else do we have going on? So, at this point, we have a huge spread on all of the Claude Code models, and I wanted to break down why this is here, what the heck is going on, where some Claude Code models are insanely expensive relatively to others, and some are insanely cheap. And the reason why you see this spread is because I've tried to break some of the math down for your use cases. So you'll see I say for some of the Claude Code tools we're a light user and in other Claude Code tools I'm saying that we're a heavy user. Let me break down my reasoning so this will make more sense.
So the way Claude Code works is that you in every, you know, 5-hour window, you can only submit so many tokens at the absolute best case on the Pro plan, the $20 a month, you can submit 45 requests per 5 hours. Okay. On the Max plan, if you know, everything goes great, some people have said that they can do up to, you know, you get 5x this number, which really means you're getting 225 requests every 5 hours. The gotcha is some of this is it's, you know, it's loose and it's up to interpretation because some people say they get close to 100. So the numbers here are very skewed because you might be sending in a long request which might take a long time which then really sucks up your usage limit because they're not counting tokens or I guess they are counting tokens but they're really counting time time Claude is running. So the numbers here are very skewed. So what I have done to be as conservative as possible is when I'm actually showing these numbers, I'm using the most conservative ones I could find. So just know when we're actually comparing numbers up top for anything related to Claude Code, just know the numbers could be half the price. So just, you know, you see 3 cents, well really it could be 1 and a half cents. I just wanted to be overconservative to make sure I, you know, wasn't telling you something that it could be when it really wasn't.
Let's go in and compare what is a light user and what is a heavy user. So let's break it up. Here is what a heavy user is. Well, you have, you know, 5-hour windows and realistically in a 24-hour day, we'll say you get these and you're, you know, you're sleeping over here. Okay. So, here's what's happening for a heavy user. Well, they most likely if they're glued to their seat all day, they have a morning shift, you know, where they're working for their first 5 hours. So, this is in the morning, they're working on that. Then they have their lunch session where they're also using Claude Code. Again, at some point they're probably going to take a break, you know, get ready for the day. So, they miss out on one of these sessions. They're probably still sleeping because you need more than 5 hours. And then if you're absolutely grinding and getting after it, you might have one extra session in a day of using Claude Code because they break it up in 5-hour chunks. So, in a realist, in a real world, you're getting two to three working sessions every day using Claude Code. And this is a heavy user.
Now, here's the difference. A light user, well, they're at their job in the morning and in the afternoon, and what they're having to do is they're coding on their projects in the evening. So, they really have one session per day to use these coding tools to get the absolute most out of them. So, what that means is they get one, a light user gets one session per day with Claude Code. And for math, just to keep it super simple and once again underestimate what the value is of Claude Code, we're going to say they get two per day. Okay.
So, using those numbers, what ends up happening, just looking at the math, if you are using a tool like Max, like if you're using the Max plan where you get 5x usage and you're a light user, we're going to say, hey, instead of 225 requests per hour, we're going to do 100. You're going to work every day of the month, you're going to get a total of 3,000 requests per month. Realistically, this could be closer to 6K, but I just want you guys to get the most conservative numbers. And on average, that's going to be 3 cents per request. And this is when every three minutes during that five-hour period, you're sending in a request. That's the average.
Now, what does that look like for a heavy user? Well, a heavy user, they're seeing a 100 messages. Every 5 hours, they have two coding sessions per day. So, you know, roughly 60 in a month. So, they're getting 6,000 requests in a month. Realistically, it could be closer to 12K. And this is going to cost them on average 1.7 pennies, but once again, it could be half. So, I just want to show you guys these numbers because your usage directly determines what value you're going to get out of these tools. That's the most important thing from that exercise. So, that's why Claude Code down here for light users is actually, you know, not that much more affordable than Cursor. So, if you're just working on the weekends and you're not or in the evenings and you're not cranking out insane amounts of code, hey, like there's really no huge benefit to go to a Claude Code bigger paid plan because Cursor gives you more features and it's going to give you more models. So, yeah, just wanted to call that out.
Now, if you're a heavy user like myself, you're going to get insane value out of these Claude Code models because, as you can see, when I'm using Claude Code heavily, where I'm doing at least two coding sessions a day, I'm paying 1.5 cents per request, which is crazy to me that I'm getting that much value for one half cent on every request. And just a quick reminder, these are with the most conservative numbers. It realistically is half. Like, I'm paying 0.88 pennies per request if you hit that more 200 cap instead of the 100 requests per hour. So it just, key thing here is if you're a power user, you basically there's nothing else that you should be using. Claude Code is the absolute beast.
Now here's one thing that I do want to recommend while we're looking at these different plans. I want to cover Claude $100 plan and a $200 plan. I personally, unless I am coding 24/7, shut the door, never leave my screen, I have never really had the need to go above the $100 plan because I'm working for a long period of time and I'm working consistently, you know, morning, afternoon, and evening. I have not had any issues with going above the $100 plan on my 5-hour limit or my week limit. So, I have not had issues yet. However, if you're an absolute beast and you're using Claude Code to where it's running all sorts of automations for you, it's basically an employee for you because you have hooks that are doing actions and those actions are making reports. Those reports are getting used by other agents inside of Claude Code. Like once you're going to like that level, you know, ninja level, yeah, you might want to go to the Claude Code $200 plan. But for like normal humans that are just using these tools for just AI development, I mean, god, the $100 plan, you can go so far with it.
Okay, so hopefully that makes sense. Now, the other case I want to cover is Cursor, which is on the Ultra $200 plan. This thing is also an absolute beast. You are going to pay 2 cents per request if it's a basic model. 4 cents per request if you're using Claude 4.5 Sonnet thinking, which is just kind of funny though because even when you're paying the $200 a month plan, the second you use like a really powerful model, you're basically back down here at the bottom of the list. So even on the $200 plan, you're still not getting that much bang per buck per request.
So key takeaways that I would say from this chart is Claude Code. If you're a power user and a heavy user specifically, definitely need to look at using it because you're losing a ton of money by using some of these other tools. Outside of that, if you are a very light user, a beginner, hey, there's nothing wrong with paying a little bit more money using some of the bottom more expensive request, you know, cost per request models and tools down here because hey, they make it easy for you to build. And if you're in the middle, hey, we're going to dive into that in the next part where I'm going to show you guys based on where you're at, which four, you know, characters you are. I'm going to help you pick the exact tech stack you should use when it comes to AI coding tools to help you figure out which one you should buy. So, let's go ahead and hop over to part number three, and we're going to help you pick the exact stack for your case.
All right, so let's dive into part three and answer the question, which tools should I buy? And the way I have structured this section out is I am going to go from like, hey, I really, really need to pinch pennies. I would love to start to use some of these tools, but I just don't have a ton of money right now to actually buy the most expensive plans. Where over here on the right, you are the developer who is just like, I got cash to blow. Just tell me the best thing to use and I'll use it. So, let's actually break this down from the more affordable all the way to the most expensive.
So, let's kick this off with what I'm calling the starter pack. So, this is the approach I would pick if I was having to help out someone who was, like I said, just wants affordability and they're willing to, you know, kind of navigate between a bunch of platforms, but they just want as much tokens as they possibly can get for free and to, you know, really stretch everything out to help them on their development journey. So, here's exactly what I would buy. I would say if you want the most bang for your buck, you want to use the Claude Code $20 a month Pro plan. You are going to get around 45 requests for every 5 hours. Then what you're going to do is the second you hit that limit, what you're going to do is you'll then rotate to using Cursor and you'll use the Cursor on the free tier. I think you get like maybe 50 requests per month. And once you know, basically what you're doing is you're just always counting a timer for 5 hours. And the second you hit a limit on that 5 hours, you're hopping to a backup tool to get more usage. So you're going to, like I said, Claude Code. Second you hit that 45 limit, boom, you're going to go to Cursor and use their free tier. The second that runs out, you're going to fall back to using WinSurf as, you know, I think they have another 50 free requests that you get per month when using their plan. So, and then, okay, it's been 5 hours. Fantastic. I'm going right back to Claude Code. Basically, the fact that Claude Code gives you a 5-hour reset timer, it's an absolute cheat code for developers who don't mind, you know, bending their schedule to make their dollars go as far as possible. And then outside of that, you could also, you know, alternate to tools like GitHub Copilot and Warp. All these different platforms give you free tokens every month. So, if you're really, you know, working on a budget, this is what I would recommend as the starter pack. Nothing else is going to take you as far as going with this approach.
Next, here is what I would recommend. If you have a little bit more money, but you're like, man, I don't want to spend $100 a month. I, you know, I kind of want to be like really reasonable about this. Here's what I would do. I would use Claude Code on the $20 a month plan. Nothing has changed here. You're still doing everything in that 5-hour window where you're going to get, you know, 40 up to 45 requests per 5 hours. And as soon as you hit a limit here, boom, you're hopping back to Cursor. And now instead of having like 50 requests per month, you have 500. So, you know, every time you hit a limit, boom, you're going to hop to Cursor and start to take advantage of that 500 requests per month. The other amazing part about starting to upgrade Cursor, which is why we want to update sooner, is because we're going to have access to the 1 million context models. So, this is where we're really going to start to take advantage of potentially using, you know, Claude 4.5 Sonnet thinking on 1 million tokens. The gotcha is even though if you do want to use that model for a monster task, only problem is you're going to have to use max mode inside of Cursor. Like, you can't even use max mode until you pay this $20. And the problem with max mode is you have to go back to like an API or like a token usage. So, even though you're paying $20 a month, well, the second you use these 1 million token context window models, you're back to paying per usage and it's not included in your $20 a month. So, it's kind of a gotcha of, you know, getting access to, you know, it's the entry fee. It's the welcome to the party, welcome to the club fee to get access to that 1 million token context window that's really not supported anywhere else. So, this is what I would do as a power stack to where yes, it's going to be the base of $40 a month plus you're going to pay whatever you had to pay for those additional tokens when you're in max mode. So, but you know, hey, 95% of your requests are always going to fall within that probably that $40 a month and only the super extra like I'm solving a super complex problem. That's when you're going to go into max mode.
Okay, now let's go up the chain to when you have, you know, a little bit more money to spend and you're using AI constantly. Well, this is the exact tech stack that I am using and this is the exact one that I recommend for most power users. What I recommend is to upgrade as quickly as you can to the Claude Code 5x Max mode, which is going to cost you $100 a month. This is going to give you, if you're a normal developer just working hard and consistently, multiple sessions per day, this is going to allow you to crank out an insane amount of code with AI and not break the bank. I know a few months ago when I was building out Shipkit, I was coding from dawn to dusk every day and I was using Cursor only. I ended up racking up a bill because I was working on nine separate projects. All the projects I included in Shipkit where I was teaching people how to build a RAG application, chat applications, a SaaS version of each one of them, SDK, I was building out all these projects inside of Cursor and I was on max mode so every time I hit enter on my keyboard I was paying 10 cents and I was using max mode a lot so over the course of a month I racked up a bill of over $3,000, which is insane because I wish I had done this exact exercise right here because I most likely would have just used the Claude Code $100 $100 a month plan and my bill probably would have been like easily half to a third or a quarter of that just by switching up which tool I used to build out Shipkit. So this is, yeah, I wish I'd known about Claude Code before building out Shipkit.ai because I would have saved so much money. So this is my current day power stack. This is what I recommend all power users to use. You are pretty much never going to hit any issues with hitting token limits and you're going to have the ability to use max mode when solving those big difficult tasks.
Now, let's go to the absolute baller of, hey, I just want the best and the greatest. I don't care about the price. I just want to use, you know, these AI tools like there's no tomorrow. Well, in that case, what I would recommend is to upgrade to the Claude Code $200 a month plan. And this is going to give you a weekly usage limit that's virtually impossible to hit. I've been over the past month pretty close to hitting my weekly limit of I've been hitting like close to like 70% of my weekly limit. And you know, potentially there are some months when I would have to upgrade to $200 a month, but that's with me coding from dawn to dusk. So, that being said, if you're coding absolutely non-stop, you're in absolute beast mode and you're just cranking out code like there's no tomorrow, the next immediate move is to upgrade to Claude Code Max mode on the 20X plan. So, yeah, this is the 20X plan.
Now, here's the gotcha. What you might be wondering is like, Brandon, why the heck are we not updating Cursor? Why are we still sticking to $20 a month? Well, we're only using Cursor in our case as strictly an IDE and as a fallback for 1 million tokens. At least, that's how I'm using Cursor nowadays. I'm not using it as much anymore for doing day-to-day changes because it was just costing me an arm and a leg. You can only pay $3,000 so often before it just takes all your money, you know, because that's what I felt like. So, yeah, 10 out of 10, you know, there's really no benefits of upgrading Cursor above the $20 a month plan. So, that's, you know, my analysis after going through this process.
But yeah, I would love to hear where you guys fall in this power stack, you know, in these different stacks of figuring out which tools you should use for your situation. If you have any questions on any of this, please let me know because I want to make sure you guys are getting the absolute most out of these tools. And just as a quick reminder, if you want to see how I am using Claude Code to feel like I'm inside of Cursor, you definitely want to grab this cheat sheet where I walk you guys through everything from how to officially set up Claude Code, how I use Claude Code with all sorts of shortcuts to make things as easy as possible. How I'm setting up my settings for Claude Code, how I have it to where I have noises ding when Claude Code is done running in the background. I put a lot of work in here to make it to where you guys can, you know, basically just spin up everything and you're just an absolute wizard when using Claude Code inside of Cursor or Visual Studio Code.
And that's a wrap for this video, guys. I hope you enjoy seeing this Claude Code versus Cursor comparison. I'd love to know which bucket that you guys are going to go with all the way from using Claude Code only to using Cloud Code plus Cursor. I'd love to hear what you guys are doing. And if you have any questions, definitely just drop it down below and I'd be happy to help. And also, just as a quick reminder, check out that free cheat sheet I got for you guys if you're looking to join the school community with over 10,000 AI developers. It's completely free. We have weekly free coaching calls. Would love to see you in there. And if you're trying to bring your AI projects to life, I definitely recommend checking out shipkit.ai so that you guys can start to build your own AI applications in days instead of months. We have coaching calls, project templates, you name it, we got it inside of Shipkit and be happy to help you with that as well if you have any questions. But outside that, I have a ton of other AI related content right here on the channel. And I definitely recommend checking out whichever video YouTube is suggesting right now for you guys. And I cannot wait to see you in the next video. Have a great day. See you.