6. How to Navigate Claude Co-work: A Complete Interface Walkthrough

Key Takeaways

Hello and welcome back. Before we start throwing real tasks at Coda, let's take a proper tour of the interface because knowing where things are and what they mean will save you a lot of confusion later. We are going to walk through every panel and more importantly, we are going to understand each one through live examples, not just theory. Let's open Cloud desktop and get into it. The first thing to notice is right at the top center of the Cloud desktop app, a mode selector with three options. Chat, co-work, and code. This is how Cloud desktop switch personalities essentially. Chat is the familiar conversational Cloud you already know. Great for quick questions, writing help, and brainstorming. Now, let's switch the mode and look at code. Code is basically the Cloud code, a powerful agent decoding environment for developers. Here you can edit the code as well. Code is the Cloud code, a powerful agent decoding environment for the developers. Here you can open your directories and start coding with small edits as well. Now, let's switch the mode to co-work. It is basically what we are focused on. It is the middle ground built for knowledge work. Think document creation, file organization, research synthesis, anything that involves your local files and multi-step execution. You can see right now we are on the co-work tab. The interface has shifted. The heading says, "Let's knock something off your list." And instead of a chat thread, you're looking at a task workspace. This distinction matters. You're not having a conversation here. You're assigning work. Now, let's go through each part of the interface systematically. The task input bar. Notice it sits right in the center of the screen when you have no active session with a placeholder text, "How can I help you today?" Once a task is running, this bar moves to the bottom of the active task view and becomes how you send mid-task instructions to Cloud. Think of it like a brief you hand to a colleague. Instead of typing, "Can you help me with my files?" you would type something like, "Go through my downloads folder, group all the PDFs by year based on their file name, or you could say metadata as well. And add one more task and create a summary. Since I have lots of data in my downloads folder, let's take the first top five. That's the level of certainty that makes Coda work shine. Also, notice two things inside this input bar. There is a work in project drop down with a plus button. This lets you attach this task to a project, which we will cover separately. On the right, you can see the model selector, currently showing Sonnet 4.6. According to your preferences, you can go with some other model as well. Then, on the left sidebar, this is your navigation hub. At the top, you have new task to start fresh. Below that, search. It basically lets you find past sessions. This is where you manage recurring tasks you have set up with the schedule command. Let's look at the left sidebar. This is your navigation hub. On the top, we have new task. This is where you start fresh. Below that, we have search option. It basically lets you find the past sessions, or basically the chat history. Then we have scheduled. This is where you manage recurring tasks you have set up with the {slash} schedule command. Then, the project. The projects tab help us basically group related tasks into persistent workspaces. Then we have ideas. On the interface, you can see that we have various built-in tasks that have been provided by the cloud itself. This is where cloud can help you brainstorm tasks to run, like write a blog post, build a web app, or automate a browser task. You can start your own as well, like by going to create button or analyze. These are the various categories that cloud offers. Now, let's go back. Below those navigation items, you'll see a recent section. Right now, it says no sessions yet because this is a fresh install. After you run a few tasks, your history will populate here, and you can click any past session to review exactly what Claude did step by step. Then you can go to the settings icon by clicking on the profile by clicking your name and avatar at the very bottom left. You You'll see your name, so click on that, then you can choose the settings option. Inside settings, Co-work is where you'll configure the global instructions, folder permissions, and more. We'll visit this in the productivity tip section. Right now, you can see the account, usage, and the capabilities. We have connectors as well. We'll cover this in a separate section. You'll see the Co-work. Here, you can edit the global instructions for your workspace. Then we have extensions as well. You can browse the extensions that Claude Co-work offers. So, let's try using them. Go back and click on the new task. Let's see. Let's pick a task that Claude is offering, organize my screenshots. When I clicked on organize my screenshots, Claude offered me a built-in prompt. You can make some specific changes if you want or just experiment with the existing one. We have other built-in tasks as well. Like optimize my week, organize my screenshots, and find insights in the file. These are not just decorative. You can actually click any of them to kick off that type of task immediately. And at the bottom, there is customize with plugins. And at the bottom, there is customize with plugins, which lets you add specialized capabilities to Co-work. You can choose any of them to experiment with. In the further videos, we will be experimenting with various plugins to perform some hands-on task. Now, let's try performing this action, but I'll make some changes to the prompt and select the desktop folder. I'll give the access. You can't really understand it by looking at an empty interface. So, let's actually give something. Here's the thing about the action log. You can't really understand it by looking at an empty interface. So, this is what we have asked Claude to do. I have given a small task. Watch what happens on the right side of the screen. The action log panel has appeared and it's already populating with the screenshots that I have in my desktop folder. Over here, it has divided that task into four subtasks. The first few lines are planning entries. Claude is analyzing the request and deciding how to approach it, which you can see in the progress section. You might see something like analyzing task structure or determining the file access approach. This is Claude's prep work before any real action begins. It's like watching someone read a recipe before starting cooking. So, over here you can see that it is reading the files. All the files that are there in my desktop folders. Next come execution entries. These are the action lines. You'll see Claude reading from your desktop directory, cataloging file extensions, counting the occurrences. Like over here, you can see the total images on desktop is 61. Each line tells you specifically what Claude touched. This is critical. If you ever see Claude reading from a folder you did not expect, say your documents folder when you only mentioned desktop, that's the moment to pause and ask why. Over here, we can see that it has performed the task and it has grouped all the files in different sections, like a group sort of things. If the task were more complex, say reorganizing hundreds of files, you would also see sub-agent entries. That's Claude spinning up parallel work streams, like having multiple assistants working simultaneously on different parts of the job. The log would show concurrent threads of activity and it might look chaotic at first, but it's intentional and efficient. Finally, a completion marker appears, usually a summary line telling you the task is done and what was produced. Over here, we can see that it has produced 37 files with 11 folders and 36 renamed files with one suggested deletion. The habit to build here, don't just let the action log scroll past. Give it a glance every 30 seconds or so during a running task. If everything looks consistent with what you asked for, great. Let it run. If something looks off, that's your cue to intervene. As you use Clowork more, you'll manage multiple tasks across different points in time. The interface reflects three states a task can be in. When you submit a task, but Clowork hasn't started it yet, perhaps because another is running or you have scheduled it for later, it lives in the queue. Think of it as a waiting room. Active task view is what you see while Clowork is working. This is where the action log is live and where the pause and cancel control appear. There is only ever one task running at a time in a standalone session. Now, the control that gives you real power over a running task. Whenever you run a task, like say, I'll say run it again. Once it starts running, you'll see the pause option over here. Pausing is your first and gentlest option. When you hit pause, Clowork finishes the individual step it is currently executing. It won't cut off mid-file right and then stop and wait. Like you'll see, hit pause and add a clarification in the task input bar, then resume. Clowork incorporates your new instruction and continues. This pause, add context, later zoom loop is one of the most useful pattern in Clowork. Then we have canceling option also. Canceling is the full stop. Use this when the task has gone in a fundamentally wrong direction and you would rather start over with a better brief. You have the option to rewrite as well or basically restart the conversation from here. This is also a good moment to mention deletion protection. Clowork has a built-in safeguard. If any step in Clowork's plan requires permanently deleting a file, it will pause automatically and show you and show you a confirmation prompt asking for explicit permission. Like over here, I'll experiment with this command. Cut. Like over here, I will experiment with this command. It says that it is working on it listing all the folders on the desktop that was curated separately. Always read that prompt carefully before clicking the allow button. Now, it has asked that there are pre-existing folders. You can either prompt the model to choose a particular folder or just choose all of them. You now have a complete mental map of the Claude work interface, the three modes at the top, the task input bar, and how to write the prompt, the side bar navigation, where to find settings, and how to read the action log as your real-time window into Claude's work. How to read the action log as your real-time window into Claude's work, and how to stay controlled with pause, resume, and cancel. In the next video, we are putting all of this into practice. We'll run your first real multi-step task, something meaty enough that you'll actually see sub-agents in action, and have a chance to do a mid-task pause and redirect. Thank you for watching, and I'll see you in the next one.