Understanding the Interface
Navigate the Cowork interface — task panel, file browser, context window, and keyboard shortcuts for faster workflows.
You’ve run your first task. Now let’s get comfortable with where things live — so you can work faster and catch problems before they happen.
🔄 Quick Recall: In the last lesson, you switched to Cowork mode and ran your first task using a specific, scoped prompt. Remember the three parts? Specific scope, clear deliverable, safety constraint.
The Cowork Layout
Cowork’s interface has a few key areas. It looks different from regular Claude chat, and understanding the layout saves you time.
The Task Panel — This is where you type instructions and see Claude’s progress. Unlike chat, you’re not having a conversation. You’re assigning work. Claude shows you its plan, asks for approvals, and reports back when it’s done.
The File Browser — Shows the contents of the folder you granted access to. You can see what files exist, which ones Cowork has read, and what it’s created or modified. Think of it as a window into Cowork’s workspace. If something looks wrong, you’ll spot it here.
Status Indicators — While Cowork is running, you’ll see progress markers. “Planning,” “Reading files,” “Writing output,” “Waiting for approval.” These tell you exactly what Cowork is doing at any moment.
Context Window: Cowork’s Working Memory
This is one of the most important things to understand, and most guides skip it.
Cowork has a context window — essentially its short-term memory. Everything it reads, every instruction you give, every file it processes takes up space in that window.
Why does this matter?
If you ask Cowork to process 50 large files at once, it might:
- Forget details from earlier files
- Make mistakes on later files
- Miss instructions you gave at the start
The fix is simple: break big tasks into smaller batches.
Instead of:
Analyze all 50 reports and create a summary for each.
Try:
Start with the first 10 reports in the folder (files starting with A-D).
For each one, create a 3-sentence summary.
When you're done, I'll ask you to do the next batch.
✅ Quick Check: Why would you break a 50-file task into smaller batches rather than processing everything at once?
Structuring Multi-Step Tasks
Simple tasks — “summarize this document” — are one prompt and done. But most real work involves multiple steps.
Here’s how to structure a multi-step task that Cowork handles well:
I need to prepare a monthly report. Here's the plan:
Step 1: Read the sales data in sales-march.csv
Step 2: Calculate total revenue, top 5 products, and month-over-month change
Step 3: Create a one-page summary document called march-report.md
Step 4: Include a comparison table with February's numbers (in sales-feb.csv)
Start with Step 1 and show me the data before moving on.
Notice what this does:
- Numbered steps keep Cowork focused on sequence
- Named files remove ambiguity about what to read
- “Show me before moving on” gives you a checkpoint
You can also just give the end goal and let Cowork plan the steps itself. Both approaches work — structured steps give you more control, while open-ended goals let Cowork show what it’s capable of.
Keyboard Shortcuts Worth Knowing
A few shortcuts that speed things up:
| Action | Shortcut |
|---|---|
| New task | Cmd+N (Mac) / Ctrl+N (Windows) |
| Cancel current task | Cmd+. (Mac) / Ctrl+. (Windows) |
| Open file browser | Click the folder icon in sidebar |
| Switch to Chat mode | Click the “Chat” tab |
You won’t need these on day one. But after a week of using Cowork, they become muscle memory.
Session Behavior — What Persists and What Doesn’t
Here’s something that catches people off guard: Cowork doesn’t remember previous sessions.
Every time you start a new task, it’s a fresh slate. Claude doesn’t know what you worked on yesterday. It doesn’t remember your preferences from last week.
Within a single session, context carries forward perfectly. You can say “now do the same thing for the April data” and it knows exactly what you mean. But close the app or start a new session? Clean slate.
This is actually fine for most workflows. You’re giving Cowork specific tasks with specific files — it doesn’t need to remember your life story. But it means you should keep your prompts self-contained. Include the context Cowork needs, every time.
Pro tip: Create a text file called instructions.md in your working folder. Put your standard preferences there — formatting rules, naming conventions, things Cowork should always do. Then start sessions with: “Read instructions.md first, then do the following task…”
That’s your workaround for persistence. And it works surprisingly well.
When to Use Chat vs Cowork
A quick decision framework:
| Use Chat When… | Use Cowork When… |
|---|---|
| You need a quick answer | You need work done on files |
| You’re brainstorming ideas | You need to process multiple files |
| You want to draft something in the conversation | You want the output saved as a file |
| You’re asking about concepts | You’re asking Claude to execute a plan |
| The task takes one response | The task takes multiple steps |
If you find yourself thinking “I’ll copy this response and paste it into a file” — that’s a sign you should’ve used Cowork.
Key Takeaways
- The file browser shows what Cowork has accessed — use it as an audit trail
- Context window is Cowork’s working memory — break large tasks into batches
- Multi-step tasks work best with numbered steps and checkpoints
- Sessions don’t persist — use an
instructions.mdfile for consistent preferences - Chat is for conversations, Cowork is for file-based work
Up Next
Now that you know the interface, let’s get hands-on with the core capability: working with files. Reading, creating, editing, organizing — the things you’ll do every day with Cowork.