This is the path with no terminal in it at all. You download one application, click a tab, point it at a folder, and watch an agent work on your files while you approve every change in a visual diff.
The app has three tabs. Chat is conversation with no file access. Cowork is an autonomous agent in a sandboxed virtual machine. Code is the interactive coding assistant with direct access to your local files. That third one is this lesson.
Download the build for your system, open the file to finish installation, and launch Claude from Applications or the Start menu. Sign in with your Anthropic account, then click the Code tab at the top center. Pro, Max, Team, or Enterprise required.
Make one empty folder called pick-test on your Desktop first. In the Code tab, select Local to run on your own machine, click Select folder, and choose it. Start small, with something you would not miss.
Before you send a single prompt, find the permission selector next to the send button and set it to Manual. In Manual, Claude proposes each change and waits. Your files are not modified until you accept. Loosen it later, once you trust it.
Send three prompts, one at a time. Explain this folder without changing anything. Then create a file. Then change it. Every path in this course uses these same three, so two tools can be judged on the same work.
An indicator appears showing what changed. Click it, read the diff file by file, and press Accept. Then open pick-test in Finder or File Explorer. hello dot t x t is really sitting there, on your own disk.
Settings live in the dot claude folder, shared with the terminal version. Project memory travels in CLAUDE.md. Updates arrive on launch, or from Check for Updates. Switch to the terminal when you want voice input or scripts.
You now have a real coding agent on your machine, with a diff in front of every change and your hand still on the brake. Next session is the same agent in a terminal window, which is faster and listens to your voice.
This is the no-terminal path. You download one application, click a tab, point it at a folder, and watch it work on your files while you approve or reject each change in a visual diff. If the word "terminal" is the reason you have not started yet, start here.
Every walkthrough in this course sends the same first three prompts, so you can run two of them and compare honestly instead of guessing.
Claude Code is Anthropic's coding agent. The desktop app is Claude Code with a graphical interface: "a sidebar for managing parallel work, a drag-and-drop layout with an integrated terminal and file editor, visual diff review, live app preview, GitHub PR monitoring with auto-merge, and scheduled tasks. No terminal required." (code.claude.com/docs/en/desktop-quickstart)
The app has three tabs. Chat is general conversation with no file access. Cowork is an autonomous background agent that works in a sandboxed virtual machine. Code is the interactive coding assistant with direct access to your local files. This lesson is about the Code tab.
Suits you if: you are new, you want to see a change before it lands, you would rather click Accept than type. Not for you if you want to script the agent or run it unattended overnight. That is the CLI path, and it is the next session.
What you will use
Go to the Claude downloads page at claude.ai/download and pick the build for your operating system. The official install steps are three lines long:
"Once the download is completed: 1. Open the file to complete installation. 2. Launch Claude from your Applications folder (Mac) or Start menu (Windows). 3. Sign in with your account to get started."
Source: support.claude.com/en/articles/10065433-install-claude-desktop
Caution: the only download domain you should ever use for this is claude.ai. Every install link in this academy points at claude.ai, chatgpt.com, or x.ai, and nowhere else. If a page, a video, or an email hands you a Claude installer on some other domain, close it.
Do this now: download the installer, run it, and open the app once. Stop there.
If clicking Code prompts you to upgrade, you are not on a paid plan yet. If it prompts you to sign in online, complete the sign-in and restart the app. Both behaviours are documented on the desktop quickstart.
Worth knowing before you go looking for a second installer: "The desktop app includes Claude Code. You don't need to install Node.js or the CLI separately."
Make the folder first, in Finder or File Explorer, so the app has something safe to point at. Call it pick-test and put it on your Desktop. Leave it empty for now.
Then, in the Code tab:
pick-test.Caution: the folder you select is the folder the agent can touch. Do not select your Documents folder, your whole Desktop, or your home directory on day one. One small folder you would not miss is the right first choice, and the official tip agrees: "Start with a small project you know well."
The app offers four permission modes in the selector next to the send button: Auto, Manual, Accept edits, and Plan. In Manual, "Claude proposes each change and waits for your approval before applying it. Your files aren't modified until you accept."
Do this now: set the mode to Manual for your first session. You can loosen it later, once you have watched it work a few times. This is the setting that turns "an AI is editing my computer" into "an AI is showing me a diff and waiting."
Send these three, in order, one at a time. Every path in this course uses the same three, so whichever two tools you try, you are comparing the same work.
Prompt 1
Explain what is in this folder in plain English. Do not change anything.
Prompt 2
Create a file named hello.txt in this folder. The file should contain the sentence: I am set up.
Prompt 3
Add a second line to hello.txt that says: I can change files too. Show me the change before you save it.
After prompt 2, an indicator such as +12 -1 appears. Click it to open the diff view, review the change file by file, and use the Accept or Reject buttons. Then open pick-test in Finder or File Explorer. hello.txt is really there. You did not type it.
Anthropic documents voice dictation for the Claude Code CLI and for the Claude Code VS Code extension. The voice dictation page describes /voice in the CLI and says the VS Code extension "also supports voice dictation with the same Claude.ai account requirement."
That page does not document voice dictation in the desktop app. So: if talking to your agent instead of typing matters to you, that is a point for the CLI path, not this one. Do not take anyone's word that the desktop app has it, including mine. Check the page above before you plan around it.
Your project files stay exactly where they are, in the folder you selected. The app's own configuration goes in your home folder, and here is the useful part: it is the same configuration the CLI uses.
~/.claude/settings.json and ~/.claude.json are shared between Desktop and CLI~/.claude.json or .mcp.json in the project~/.claude/skills/project-root/.claude/worktrees/, configurable in Settings, then Claude Code, under Worktree locationCLAUDE.md files in the project itselfSource: code.claude.com/docs/en/desktop
The desktop app auto-updates on launch on macOS and Windows. To check by hand:
Source: code.claude.com/docs/en/desktop
Anthropic's desktop documentation does not publish an uninstall command for the app, so remove it the way you remove any other application on your operating system: drag it out of Applications on a Mac, or use the normal Windows app removal.
One documented detail matters if you want it gone completely. The setup page warns that "The VS Code extension, the JetBrains plugin, and the Desktop app also write to ~/.claude/. If any of them is still installed, the directory is recreated the next time it runs." So uninstall the app first, then remove the configuration:
rm -rf ~/.claude
rm ~/.claude.json
Caution: that deletes your settings, your allowed tools, your MCP server configuration, and your session history. Only run it if you mean it.
Switch to the Claude Code CLI when any of these becomes true:
/voice is a CLI feature today.You do not have to choose forever. Anthropic documents running both: "You can run both simultaneously on the same machine, even on the same project. Each maintains separate session history, but they share configuration and project memory via CLAUDE.md files." And it goes the other way too: run /desktop in the terminal and "Claude saves your session and opens it in the desktop app, then exits the CLI." That command is available on macOS and x64 Windows when you are signed in with a Claude subscription. (code.claude.com/docs/en/desktop)
The Code tab is open on your pick-test folder, you sent all three prompts, hello.txt exists on your Desktop with two lines in it, and you approved the second change yourself in the diff view. That is a working agent, on your machine, with you still holding the brake pedal.
Template included: Members get a one-page Claude Code desktop card: download link, the Code tab path, the permission modes side by side, the three comparison prompts, and the update and uninstall steps. Find it in your member dashboard under Templates.
Short on time? Agentic Quarks will implement this end to end, wired into your stack, and hand you the keys.
One full walkthrough per tool and per surface, so you install the right one once
5 sessions, 100 minutes, 5 free.
One payment, no subscription, and a 14-day money back you trigger yourself from your account page.