Full transcript of “Amazon’s Kiro AI IDE: A Real Copilot & Cursor Killer?”. Click a timestamp to open the video at that moment. Need another video? Use the YouTube transcript generator.
0:00 Hi guys, welcome back. And today in this video we're going to talk about yet another AI IDE but this time this IDE is developed by Amazon not from Google or from Microsoft or even some other third party companies. This is a big company Amazon which has built this IDE. It's called Curo. And we already have seen in our exotition YouTube channel about the Amazon developer Q which is an AI based CLI tool which the cloud code does or the uh Google's uh Gemini CLI does. That's exactly what the Amazon developer Q does as well. But this one the Kuro is going to be competing right straight with the the VIP coding IDE like cursor IDE or Vinserf or Microsoft's GitHub
0:48 copilot. So all of these are going to be destroyed with this particular IDE. This is quite amazing and I've seen the power already and I'm going to show you that power very quickly and then we'll jump in right into this particular tool in an demo. So this tool the Kuro IDE actually does the AI coding with a specdriven development as opposed to directly where you just go and type things and ask the questions and then start writing the code or do the development from there. I know in the cursor ID as well we can create a requirement MD file and then we can put all our uh requirement that we wanted to by asking the the questions in the prompt and then it saves in the MD file and then from there we can start using it as an memory for the for the uh cursor uh ids agent to work from there.
1:37 But this one is even more fascinating because it does everything by default with a specdriven development. And they have also got another feature called as the uh the agent hook which does even more great things. So as you can see in this particular demonstration anytime that you ask any questions like can you do this operation for me like an e email opt-in form then it is going to go and start doing the requirement for you. And once the requirement is done, it then goes to the design phase where it draws the diagrams and things and then also puts the detailed information of how you are going to be designing this particular operation. And if you're happy with this design and if you wanted to even make any modifications, you can do it. And then from there, it goes to the task list where it's going to show you all the task which is required for
2:23 you to perform for every single uh design operation that you have created. And then you can also run every single task pretty much like a checkpoint and then you can roll out roll back to the checkpoint like how you do from the cursor ids and things. So these are the amazing feature which is already standing out from the rest of these vibe coding tool. And then you can see that they have got something called as agent hooks which is pretty much like a trigger where you can set up like uh if you make any change in the project structure then go and update the documentation if you set that particular trigger then that particular trigger is going to run for you automatically every single time while you make any of the change in the application. So this is a kind of one single trigger that you can do but you can set many different triggers and every single time you make the change the agent hook is going to
3:09 listen and it's going to update things for you. So this is also yet another standout feature from the uh rest of these uh toolings like cursor IDE or uh win server for that matter. So I have already tried this exact same thing with my MCP server that we have got in the GitHub which is uh this one. If you have not really tried it, I highly recommend you to go ahead and try this u uh this MCP playright. This is something which you can use for the uh playright automation using the MCP server and I have built this mostly by hand coding and then I started using cursor ID to make the development even more streamlined and faster. But now with this particular IDE which is nothing but the KO IDE it has gone a level further. So this is what I have already tried out before recording this. It has just
3:55 released like 8 hours before and I just wanted to show you the first glimpse of how it actually is looking like. You can see that over here it has now created a requirement documentation for me. Basically I asked a a requirement for it to be created. Uh like can you go ahead and uh analyze what is the industry standard missing feature in this MCP server. uh not exactly like playright tool but an MCP server which holds the features like SSE support and other things that you wanted to do and then can you analyze the project first and then tell me what are these uh features that I'm missing in this particular MCP server well I told this it just jumped right ahead and started uh doing all the analysis of the code which what kurser and vinsurf and github copilot does already I'm not going to blame them but
4:42 over here it is going to write all the requirement document like an MD file over here Right. This is you can do the same thing exactly with a cursor IDE as well. But the good thing is you now have an ability to see everything is stored in an like a memory fashion. And then it goes to the design phase over here where it is going to also draw a mermaid diagram like an architectural diagram like a highle architectural diagram to see what exactly the things are going to come out from this particular uh from this particular requirement. It's also writing a small code snippet for each and every requirement like for example over here it has designed the architecture for the core components and then it is writing the core components what are they like a SSE server component uh and the uh logging system
5:29 component and there is also a rate limiter components and monitoring system component. So it's also putting all these components for me there uh in a TypeScript format which is going to be used while it is going to do the development. So this is the design phase right? So it's going to put everything over there and then it is going to create a task list for me where it is going to put all the task mapping the requirement pretty much like what an user does in the requirement traceability. So it does the requirement trace for every single requirement over here. In fact you can just go to any one of the requirement over here and do a control T I guess or ctr L. I'm not pretty much sure what exactly is the uh shortcut. It will take you directly to the chat window where you can also ask
6:14 what you wanted to do. And guess what? Once you wanted to once you're happy with this particular requirement. Oh yeah, this is one the the the Windows uh the command L over here. So once you hit command L over here, you can run that particular specification and then you can also start uh doing a vibe coding if you wanted to and then you can also write the specification if you wanted to. that that's how you can actually see you can improve the specification and then start creating the task or updating the task from there which is even more insane and you see that now I have got this particular task over here which I want to implement if I'm just going to say okay now implement the comprehensive logging system I'm just going to hit this particular start this task and you see that now the task is in progress and the kirao is jumping right in and then it is going to start building things for
7:01 me it's pretty much like you have a specificationdriven development where you have got all the specification clearly written what exactly is going to be built for every single prompt. This is actually a prompt, right? So, it's going to put everything in the prompt and we have more control over what exactly is happening every single time. Because while I work with cursor ID, the only thing which goes very out of control is that it keeps writing every single code and sometimes I feel like some of the code is completely wrong or maybe I need to refactor that code. Then I go to the checkpoint of the cursor and then I revert everything back if I feel like some of the things are not correct and then I have to redo and ask the the the cursor IDE to write me exactly what I'm actually looking for which is kind
7:47 of a babysitting work but now this has been resolved with this with this curo because it it is telling me exactly how it has to be done which is even more amazing and you can also see that now it is trying to do the task for me and because I'm using the free version And I'm also using the the cloud sonet 4.0 over here. Uh it is going to take some time because it says that currently the execution is ceued. Uh and if I'm going to open the task over here, look at that. Now it's going to show me what are the task being executed every single time. And it it also shows me what exactly the task is being executed. So we can see there are two task which is currently being executed and they are in Q as well. And we can see in the list which is amazing. And again, I'm using the clouds on 4 here. And because this is a free version until uh this
8:34 particular preview is going to be going to the actual production, we are going to have the ability to use the most powerful models for free of charge for now. So I'm just using as much as I can by this time. You know what? This is better. So this is this is amazing. This is another feature which I really like about this particular Curo. And the last thing which I wanted to show you which differentiates Curo from the rest of these IDEs is nothing but the uh agent hook. Look at that. Now because I'm making a change in the code, it's also doing a documentation sync for me because I have created this particular agent hook which I told over here that listen to the TypeScript source file and configure changes to automatically update project documentation in the readme file and in the docs folder
9:19 because within our uh MCP server we have got the uh docs in the uh docosurus and also we have got the readme file. So I wanted both of these to be updated every single time while I make any change in the application. So that is also going to happen for me uh during the process of the change uh in the application. This is amazing. So code development is going to happen and at the same time there is going to be change in the documentation. So we can see this is one another way that we can handle this. Not only that, if you wanted to make some changes in the versioning or if you want to make the changes maybe uh in the diagram or the requirements somewhere or maybe if you want to make some changes uh in some other configurations if you make any change in the code then you can
10:05 all set up everything in this particular hook for you. That way you don't really necessarily have to do this every single time and then you remember that you know what I missed this versioning changes or maybe the documentations or maybe I I missed this and that. So those things can be taken care by this um this curros agent hook. So this is amazing and I have literally not tried this agent steering so far. So I have no idea what that is. But they also have got this MCP server feature. So you can use the GitHub uh MCP server or of course this play MCP server or the database MCP server everything you can configure it over here and then you get the same ability of the MCP servers along with this particular IDE. This is amazing. This is very clean, neat, more refined. They have learned from the rest of the
10:50 uh the wipe coding tools and idees. They know how to make this thing more better. So that's the reason why I think it is even more refined from the rest of these uh rest of these tools. I don't know for some reason it is still in the execution queue uh and this guy is still implementing uh changes for me. But uh it usually it's going to be quite faster and for some reason it is slower demo time. But I could see that this is quite good. Uh but the task is still Q. But it is going to take some time to to implement it. But I can see this is the potential of this uh this particular tool which makes or maybe set out itself completely apart from or differentiate itself from the rest of these tools like cursor uh vinservef or github copilot because this tool is uh is looking
11:38 pretty good and I'm going to try everything uh in this particular tool and I'm going to see how exactly it's going to improve my workflow uh while I'm going to use this but I have already seen some of the benefits of this particular ular tool because you see that all these code is actually been written by this guy. I have not written even a single line of code so far. So uh you see that all these code is written by this tool and you can see the number of lines of code that it has written. It's it's pretty much amazing and you may ask like karthik this is exactly what cursor does. Of course it does. But here I think I feel like it I have more granular control over what every single task is going to be doing and what exactly is happening and how the
12:25 requirements are clearly being defined. And if I want to see if I make any change or updation if I wanted to do it I can also do that. Same thing I can ask it uh ask over here in the Kuro uh Kuros chat it's going to be done for me over there. That's it guys. That's about this particular uh Kirao tool and this is quite awesome. Let me know your thoughts about this Kirao IDE but I think this is quite amazing. Once again thank you so much for watching this video. Catch you in the next one. [Music]
