Transcription
Holy smokes. I did not expect to be making this video today. But if you've been using claw code in the terminal, inside VS Code, or anything like that, and you've been waiting for a reason to switch to the desktop app, then today might be that day.
So, last week Claude dropped an update. And while the announcement focused on running sessions in parallel and reviewing things without leaving the app, they forgot to mention, or they buried, two features that are actually going to change how most people use Claude inside the desktop.
So, the real issue before was that the desktop app was always super one-dimensional. So, remember back to the days of ChatGPT, chatting back and forth with one single task. Everything lived in an isolated workspace. Then came along Claude desktop. It brought Claude co-work, as well as Claude code. But even then, everything was designed to be a back-and-forth chat with individual tasks. So, you had no visibility into your wider business context, no connection between your sessions, and no sense of progress across projects. These are all things you absolutely need when you're working on a real business problem. And it's exactly the reason we built our own agentic operating system in the first place, which is effectively a pre-built file structure that stores your context, loads it in only when you need it, and is designed to improve its own outputs every single time you use it, just by managing context in the right way.
But you couldn't properly use that in the desktop app before because the files were invisible. Now, you can actually see and interact with them. So, instead of those isolated chats, you've got the framework to carry out structured, complex projects. Each client can have their own folder. You can move into that workspace, run tasks inside the workspace, and keep everything tied with the right context. And because of that, you can plan properly, track progress, and actually start to improve your outputs over time. So, effectively, the same system that used to require the terminal, due to its flexibility, now can work natively on desktop. And I'm stunned that it's happened so quickly, to be honest.
So, let me show you exactly the features that were buried in there that I think are going to be game-changers and might just move you off the terminal today, directly into the desktop app. I've also noticed a few limitations, which we'll touch on as well.
So, firstly, this was announced: multiple sessions. So, you can see that we have multiple sessions running at once. So, you can see a holistic view of all of your sessions. It's no longer chats on the left-hand side. It means you can work on actual projects where you have multiple tasks or multiple sessions pertaining to a single project. You can also see how they all connect.
Secondly, let's jump into one of these. Say we have a plan. We used to have to view the plan down here. But Claude has quietly launched this sidebar here that has options to actually view the plan inside markdown. So, say we want to view the plan, we now get the markdown rendered and formatted in this nice plan view that we can expand and collapse. We can see exactly what plan Claude set out for us, and we can also therefore execute against the plan inside the chat without having to go all the way back up and viewing the plan. You can now track progress side by side.
Secondly, split view. So, having a holistic view of a given project where you have multiple sessions is super important. And what's more important is actually being able to view those in a split view. So, say, for example, I wanted to view two of these sessions together, then I'm absolutely able to do that. I just drag and drop, and I can also open up the plan on this here, and they are completely independent windows. So, I can have multiple windows running. Say I want to run a third session, I can actually run that in there, and we can run up to four sessions at once, adjusting how we please and using the real estate on the screen how we want.
So, we can go into the plan and we can go into the files. So, previously, similar to the plan, we never had any of our outputs rendered at all, and this made it really difficult. This is also a pain in working inside VS Code because you had to go and render things like PNGs elsewhere without a preview window. Let me close these PNGs down. We're not only able to view the plan, but let's say we want to view a previous output. We go into the same window. We can go into files, and as long as we have an operating system in the back end telling it where to put output files, we keep ours in projects, for example. We can go into any of these output files, especially if they're markdown, and actually see the full preview of that file rendered in markdown on the right-hand side. Yes, it's not perfect. It doesn't render exactly, but these are things that Claude will fix. But you can effectively start to see how they're going to build out an ability to see your output files all in one place, rendered in markdown, and then able to actually go in and directly edit these markdown files.
Now, we've been literally building out a UI to do this exact thing: see holistic views of projects, edit markdown files. And I'm quite amazed that Claude has got to this phase already. But one of the most important things, although we jumped into a file that I haven't mentioned yet, is this file structure. So, previously, we were never able to see this file structure at all. So, if you were building your own agentic operating system in the back end, we weren't able to see these files and the file structure in the back end. We weren't able to actually go into our Claude MD and make edits. We weren't able to display our outputs that are being put into projects. We weren't able to add in and read our context about the agent or our brand context where we have our positioning, our voice profile, all of that stuff. I can now preview and actually update in the Claude desktop app.
And it goes without saying that now, when we're working inside Claude Code, we can actually choose when we start a new session which folder we want to start in. So, previously, we were kind of working in an isolated environment with zero context. Now, you can work inside your business operating system, your agentic OS, and you can pull all of the context from that specific folder. And the reason that's super important is because now you can actually work directly in an individual client folder. You might have a bunch of context that spreads across all of your clients, and then specific client brand context that sits within each job. You might have specific projects that relate to those. And as long as you've got these rules defined in your agentic operating system that sits behind the desktop app, you're able to effectively use it now in a really clean way with a UI inside the Claude desktop app. And all of this building out this agentic operating system, your own operating system, helps you achieve 10x results without having to look at another terminal at all.
Now, there are a few downsides that I've noticed, but I'm sure they'll be fixed soon. The first is, if you're looking at, for example, let's say we've got the Excalador skill that's outputting PNGs. If we try to render that, it's just going to give us the PNG as code rather than actually rendering the output. That's something I'm sure they'll fix super quickly.
The second downside is that we can't still see our skills folders. So, our hidden files like Claude, we're not able to actually see any of the skills inside here and edit them directly. We're still forced to actually go to the file structure or open it in VS Code and actually make those changes directly in the back end to our skills. That's something I'm sure will be also fixed soon, or at least the ability to view the skills inside here if they're in an underlying folder.
And the third and final thing, which is a bit harder to get round, is the .env files. So, the hidden environment variables where you keep all your credentials for your API keys. Right now, there's no ability to access that and change those. So, you have to go to the underlying file structure or open it up in VS Code to actually change and add credentials. However, if you're working in the agent OS repo, it's still able to access those credentials and use them in API calls, etc. So, it has access. You just can't change them yet in the Claude desktop app.
So, given all these updates, I can actually now see myself starting to fully move away from VS Code. And I'm curious if there's any features I've missed, anything that you think is still holding you back from jumping to the desktop app. Drop a comment below. Subscribe if you found this valuable.