If you’ve been using Claude via the website for your day-to-day AI tasks, there’s a good chance you’re only scratching the surface of what’s now available. Anthropic has quietly shipped one of the more significant upgrades to hit the AI assistant space this year — and most people don’t even know it’s there.
Three Modes, One App
The Claude Desktop app — available on both Windows and macOS — now brings together three distinct ways of working with Claude, all in one place:
- Chat — the Claude you already know: conversational, fast, great for questions, writing, brainstorming, and analysis
- Cowork — an agentic mode where Claude can actually do work on your computer with access to your local files and apps
- Code — a full development environment powered by Claude Code, designed for developers who want AI assistance directly in their workflow
Open the desktop app and you’ll see all three as tabs at the top of the window. Each is its own mode with its own purpose. Chat is what most people are already doing. Cowork and Code are where things get genuinely interesting.
What Is Cowork, Actually?
This is the one worth spending time on, because it’s a genuine step change in how AI can assist you.
In regular Claude chat, you upload files and interact one message at a time. Cowork can directly read and write to your local folders with much more autonomy — it creates plans, executes multi-step tasks in parallel, and works independently while you step away. It’s designed for complex, long-running work rather than quick Q&A.
Launched in January 2026 as a research preview, Cowork moved to general availability in April 2026. It takes the same agentic architecture behind Claude Code — file access, multi-step execution, tool integration — and wraps it in a desktop UI that anyone can use. No terminal required.
Some real-world examples of what Cowork can handle:
- File organisation — “Organise my Downloads folder by type and date” and Claude will sort hundreds of files into categorised folders, proposing a plan before touching anything
- Document processing — Drop receipts in a folder and ask Claude to create a formatted expense report
- Research synthesis — Pull information from multiple sources and compile it into a structured report
- Recurring tasks — “Pull my metrics from the analytics dashboard and drop them in the weekly report template every Friday” — and it actually does it, on schedule
The key principle: you describe the outcome, Claude figures out how to get there, shows you the plan before acting, and loops you in at any significant decision point.
How It Handles Your Files and Privacy
This is the question I’d want answered before handing an AI access to my file system, so let me address it directly.
Cowork runs in an isolated virtual machine (VM) on your computer with controlled file and network access. Your documents and data stay local — they don’t leave your machine for training or cloud storage. You grant folder-level permissions, meaning Claude only touches what you’ve explicitly said it can touch.
When Claude does interact with your screen (via the Computer Use capability), it asks permission before accessing each application, and you can stop it at any step. Deletion protection is built in: Claude requires your explicit permission before permanently deleting any files.
For enterprise deployments, Team and Enterprise admins get role-based access controls, spend limits, and OpenTelemetry monitoring across the organisation.
The Mobile Piece
Here’s where it gets clever for people on the go. Your conversations, projects, memory, and preferences sync across all devices when you’re signed in.
You can share a task from your phone and Claude can get started in Cowork — so you can kick off a piece of work from your phone during your commute and come back to it completed on your desktop. On mobile, Claude can also connect to device-specific apps like Maps and iMessage.
The limitation to be aware of: the desktop app needs to be running for Claude to complete Cowork tasks. If your computer is asleep, Claude won’t be able to work. For long-running cloud sessions in Claude Code, that restriction doesn’t apply — those continue even if you close the app.
What This Means If You’re Running a Business
For solopreneurs and small business operators, the use case is compelling. Knowledge work — the stuff that eats hours every week — is exactly what Cowork is designed to automate:
- Processing and categorising documents
- Generating recurring reports from templates
- Organising and summarising research
- Synthesising notes from multiple sources into a single deliverable
Cowork is available to paid Claude plan subscribers on Pro ($20/month), Max ($100–$200/month), Team ($30/user/month), and Enterprise. Pro users get the same features but may hit usage limits sooner, since complex multi-step tasks are compute-intensive. If you’re planning to rely on it heavily, Max gives you significantly more headroom.
The Code Tab — For Developers
If you’re on the technical side, the Code tab is worth your time. It’s the full Claude Code experience integrated directly into the desktop app — no terminal juggling required. Each conversation is a session with its own chat history, project folder, and code changes. You can run multiple sessions in parallel, monitor PRs, and review diffs without leaving the app.
For longer-running or resource-intensive tasks, there’s a Remote mode that runs on Anthropic’s cloud infrastructure — so your local machine doesn’t need to stay on and you can check progress from any device.
Getting Started
If you haven’t downloaded the Claude Desktop app yet, head to claude.com/download. The install is straightforward on both Windows and macOS.
Once you’re in:
- Sign in with your existing Claude account
- Look for the mode selector tabs — Chat, Cowork, Code
- Start with Chat if you’re already familiar with Claude via the web
- When you’re ready, switch to Cowork, grant it access to a test folder, and give it a simple organise or document task to see how it works
My suggestion: before you hand it a high-stakes folder, run it on something low-risk first. Watch how it proposes a plan, check you’re comfortable with what it intends to do, then let it run. Once you’ve seen it work a few times, it becomes a very different tool to a chat window.
The three-tab model — Chat, Cowork, Code — is the clearest sign yet that Anthropic is building toward something closer to a proper AI operating environment than a chatbot. If you’re still treating it as “that thing you paste questions into”, it’s worth taking another look.
Interested in how tools like this could fit into your workflow or your team? Reach out — happy to walk through it.