Claude Code Desktop App: a Tour for Non-Programmers
The Claude Code desktop app brings parallel sessions, visual diffs and live app preview to macOS and Windows — no terminal needed. Install guide and tour.
Evgenii Arsentev · PhDReviewed for accuracy by Evgenii Arsentev, PhD · 2026-10-14
The Claude Code desktop app is Claude Code with buttons instead of a black terminal window: you pick a project folder, type what you want in plain language, and review every proposed change as a visual diff with Accept and Reject buttons. It runs on macOS and Windows, includes everything it needs (no Node.js, no CLI install), and is the gentlest way I know to start building with AI.
I recommend it to every non-programmer who emails me 'where do I even begin?'. The terminal version is powerful, but the desktop app lets you see what's happening — and seeing is what builds trust in the first week.
What is the Claude Code desktop app, exactly?
The Claude desktop app has three tabs: Chat for regular conversations (like claude.ai), Cowork for autonomous background work in a cloud VM, and Code — the one we care about — for software development with direct access to your local files. In the Code tab, each conversation is a session with its own chat history, project folder, and changes. The sidebar lists your sessions, and yes, you can run several in parallel: for Git repositories each session gets its own isolated copy of the project via Git worktrees, so experiments never trample each other.
How to install Claude Code on your desktop and start the first session
Download the installer from the official docs or claude.com/download — there's a universal macOS build, Windows x64/ARM64 installers, and a beta .deb for Ubuntu 22.04+ / Debian 12+. Linux support is new and still in beta. On Windows, you only need Git for Windows if you turn on isolated worktrees for parallel sessions — a plain local session doesn't require it anymore.
Does it work on Mac and Windows?
Yes on all three. macOS gets a single universal build (Apple Silicon and Intel); Windows has separate x64 and ARM64 installers; Linux (Ubuntu 22.04+, Debian 12+) has a beta .deb package installed via apt. The only platform-specific catch: on Windows, Git for Windows is only required if you use isolated worktrees for parallel sessions, not for the Code tab in general; on Linux, Computer Use and voice dictation are not yet available in the desktop beta — use the CLI for those.
First session in four moves
- 1Install, launch Claude, sign in with your Anthropic account, and click the Code tab at the top center. If it asks you to upgrade, you need a paid plan first.
- 2Choose where Claude runs: Local for your machine (pick this first), Cloud for cloud sessions that survive closing the app, or SSH for a server you manage. Click Select folder and choose a small project you know well.
- 3Pick a model from the dropdown next to the send button, leave the permission mode on Manual, and type a task: 'Find a TODO comment and fix it' is a classic opener.
- 4Review what Claude proposes: a diff view shows exactly what will change in each file, with Accept and Reject buttons. Nothing touches your files until you accept — and if you reject, Claude asks how you'd like to proceed instead.
Is the desktop app free to use?
The download is free, but the Code tab requires a Pro, Max, Team, or Enterprise subscription — same requirement as Claude Code everywhere else. There's no separate 'desktop' fee, and cloud sessions you start from the app count against your normal plan limits with no extra compute charge. If clicking Code prompts you to upgrade, that's the subscription check talking.
Claude Code desktop app vs the terminal: what can it do that the CLI can't?
Quite a lot, and this is where it gets fun. Panes: arrange chat, diff, live preview, integrated terminal, and a file editor side by side, drag-and-drop. PR monitoring: after you open a pull request, a CI status bar appears with Auto-fix and Auto-merge toggles. Side chats: press Cmd+; (Ctrl+; on Windows) to ask a question that reads the session's context without derailing it. And Dispatch: message a task from your phone, and it can spawn a Code session on your computer while you're out.
Two additions I now use constantly: `/diff` opens a live side panel instead of a modal — it lists every changed file with line counts, refreshes as you edit, and lets you select specific lines to attach to your next prompt. And any pane — diff, terminal, whatever — can be popped out into its own window, dragged to a second screen, then docked back when you're done.
Can Claude preview and test my app?
Yes — this is the Preview feature, and it's the one terminal users miss most. Claude starts your dev server in an embedded browser and verifies its own changes: it takes screenshots, clicks around, and fixes what it finds. Instead of you switching to a browser to check whether a button works, Claude checks for you and reports back, all inside the app.
On macOS, there's now a background computer-use mode (beta, Pro and Max plans): Claude can see and act in apps you've approved while you keep working on something else — so it's not blocking your screen while it runs a long task.
Press Cmd+N (Ctrl+N on Windows) for a new session and give it a different task. Each session works in its own Git worktree — an isolated copy of your project — so two Claudes never overwrite each other's files. This is the feature that quietly doubles your output.
Coming from the CLI — or heading there?
Desktop and CLI run the same engine and share configuration: CLAUDE.md memory files, MCP servers, hooks, skills, and settings all work in both. You can run both at once, even on the same project. And there's a one-command bridge:
/desktop
Run this inside a CLI session: Claude saves the conversation, opens it in the desktop app, and exits the terminal. Available on macOS and on x64 Windows when signed in with a Claude subscription.
/resume
Run this inside the desktop app to search your past CLI sessions by title, folder, or branch, preview the conversation, and continue it right there. It's the reverse of /desktop — the bridge going the other direction.
Practical kicker: install the app tonight, point it at the messiest folder of half-finished ideas you have, and ask Claude to 'create a CLAUDE.md with instructions for this codebase'. You'll watch it explore your project in the diff view — and you'll have onboarded yourself and the AI in one move.
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →References
Selected peer-reviewed papers and preprints on the AI research behind this topic.
- [1]Rose E. Wang et al. (2024). Tutor CoPilot: A Human-AI Approach for Scaling Real-Time Expertise. arXiv:2410.03017
- [2]Mingchao Jiang et al. (2025). SIMCOPILOT: Evaluating Large Language Models for Copilot-Style Code Generation. arXiv:2505.21514
- [3]Zhenran Xu et al. (2025). ComfyUI-Copilot: An Intelligent Assistant for Automated Workflow Development. arXiv:2506.05010
- [4]Chee Wei Tan et al. (2023). Copilot for Xcode: Exploring AI-Assisted Programming by Prompting Cloud-based Large Language Models. arXiv:2307.14349
- [5]Ziyang Luo et al. (2023). WizardCoder: Empowering Code Large Language Models with Evol-Instruct. arXiv:2306.08568
- [6]Can Xu et al. (2023). WizardLM: Empowering large pre-trained language models to follow complex instructions. arXiv:2304.12244
- [7]Md. Rafiul Biswas et al. (2023). Can ChatGPT be Your Personal Medical Assistant?. arXiv:2312.12006
- [8]Weihao Gao et al. (2023). OphGLM: Training an Ophthalmology Large Language-and-Vision Assistant based on Instructions and Dialogue. arXiv:2306.12174

Author
Evgenii Arsentev
PhD · Chief Executive Officer, digital health
Reading is a good start
Want to actually build this?
Guides explain. The free course transforms — personalized, gamified, and built to get you shipping fast.
Start the free course →