Key features of the NEW Claude Code desktop app you MUST know!

You can use Claude Code in the CLI in the terminal user interface, which is probably what most developers do, but Anthropic also has a desktop app, Claude Code as a desktop app. Technically, it's actually not just Claude Code. There also exists the Claude Chat, which is like Chat GPT, just in the desktop app. And there is Claude Co-work, which is a general work assistant, but there also exists the Claude Code app. And that's actually not just the CLI wrapped into a desktop app, but it's a really useful standalone app, especially since the latest update. And no, this video is not sponsored by Anthropic. I just think this new version of this app is pretty useful, which is why I'll give you a walk-through. If you want to learn way more about Claude Code, about this app, and about the CLI mostly, I got a complete course on that. It's fully updated for this new look of this new desktop app. I did that this morning. It's already live, and that may be useful to you. So, what's in this app? Well, this app has pretty much the same idea you might know from the Codex app or Cursor, there the Cursor Agents app, if you use that. You can have multiple projects in there, and in each project you can have multiple sessions. You can start a new session. You can then pick a project for this session. You can pick a Git branch. It has a built-in Git worktree feature. And in case you don't know what Git worktrees are, the idea is that you have your main repo, and Git worktrees are an official Git feature, which creates a clone, a local clone of this repo. So, on your system, the project is kind of copied, not all the files, just the Git managed files though. And then you have that copy, and you can work in the main repo and the copy simultaneously, have multiple agents running across these different worktrees, so that they don't clash with each other. That's the idea. And then, once you're done, you can merge your worktree changes back into the main repo. That's the idea, and if you check this box, the desktop app will create such a worktree for you automatically for whatever task you're handing off to it. So, that you could have multiple agents working in the same repo without crashing, as I mentioned. You can also start cloud tasks from in there. I covered this in more depth in my course, but the idea is that if you switch to a cloud environment here, you have to set that up in the Claude Code cloud web interface first. But if you have that, you could change the primary color to red or something like this. And now, if I would submit this, this would not be executed locally, but instead in the cloud, but of course local is always an option, it's the default essentially. You can manage your permissions down here, including bypass permissions, if you enabled that in the settings. You can switch to plan mode. You can add additional context. And you also got slash commands available here. If you type slash whatever, you can essentially run all the key commands you know from the CLI as well, including your skills. You can invoke them from there, too. If you don't want to hope that Claude Code decides to load them, you can just actively invoke them from here, of course. You could also use voice dictation, and of course, you can also switch easily to different models. Now, that's all nice. I mean, you can all find that on your own. I guess still nice to see it. What is interesting is that if you have an active session, or a session that's already complete, you have this button here in the top right corner. And that is the most interesting button when doing local development with the Claude desktop app, I would say. Because here, you can for one view your plan, if you did use plan mode. Here, I didn't, so I have none. When it's currently working, you can see your ongoing tasks here, too, so the tasks it's going through. But if it's done already, you can, for example, use the preview feature to have it launch a dev server for you, and show you a preview of your website, if you are working on a website here. You have to set it up once, and if you click this button, it actually uses the Claude agent, the Claude models to write a Claude launch. json file for you, which will contain the setup instructions, or the setup configuration, I should say. So, that's a one-time step. And once it's done here, you can ask it to go ahead and start a dev server, and start that preview. And then, it will indeed show you a preview of your site here. Now, what is interesting is not just that you have a preview here, which can already be nice. You can prompt the way here, and see the changes live here to the right. That can be nice. Uh but, the best feature here is this button. This one. I mean, this one is interesting, too. You can switch between mobile and desktop view here. You can also switch between the different modes, if your app supports it, different light and dark modes. But this button here is great. It allows you to select an element, and then this element is added as context here. And you can then, of course, base your prompt on that. So, if you know that you want to change something about that

element, change the text to just welcome, you don't need to describe in which file it is, and hope that it finds it. No, you can just select whatever you're not happy with, and that will be added as context here. Of course, you can also delete it. And that can be a nice feature. And I know other tools have that, too, but the Claude Code desktop app now has it as well. Now, besides that, you also have a diff view here, and that is also pretty nice, and a feature I already know from the Codex, and also other apps. Here, you can easily scroll through your files, and see all the changes, everything that is uncommitted essentially here. And you cannot just view them, which is already nice, because reviewing the code is important, unless you're of course white coding, and don't care about the code at all. I do care though. But this is nice for code reviews, and you can also drop comments here. You can say something like remove this, or whatever. And if you submit this comment, it's also added as context here. So, now you could send it like this, or add some extra instructions. We don't want to use custom variant, whatever. You can send that. And again, you don't have to describe to which file you're referring. Instead, you can just comment in line, and then base your prompt on this, which of course can be more convenient. Now, outside of that, you also have a terminal. So, that is essentially the same kind of terminal you, for example, also have in Visual Studio Code. So, you can use the integrated terminal there. You also have it here in the Claude Code desktop app. And therefore, altogether, I think this is pretty nice. Of course, in the end, let's be honest, it is mostly what we already saw from other apps like Codex, for example. But it's a nice glow-up of the application. I like the look and feel. It's really fast and snappy in my first tests. And I think if you are into this kind of TVO development, if you prefer that over a CLI, this is definitely a great improvement over the old Claude Code desktop app. And as mentioned, I already covered it in my course. There, I also explain routines, dispatch, and much more. But hopefully, this quick walk-through also was helpful.