Skip to main content
Applied AI for Management

Scrum Masters and Project Managers: How to Install Claude Code When You Have Never Opened a Terminal

Installing Claude Code is one line. People get stuck on the four decisions around it: which way in, which plan, which window and which update path. Answer those four and the install takes minutes, with or without a terminal.

A dark desk at dusk with a laptop showing a terminal on the left and a monitor showing an app window on the right. Two thin threads of light run from both screens to one open folder of handwritten standup notes in the centre.

Installing Claude Code is one line. That is not where people get stuck. They get stuck on the four decisions around that line, which nobody mentions because they look obvious to anyone who has spent ten years in a terminal. If you run sprints rather than servers, this is the version written for you: four questions, the exact step each answer leads to, and how to prove it worked. Everything here was checked against Anthropic's setup documentation on 2 October 2026.

The clearest framing we have found comes from a guide AxonBuild wrote for exactly this reader, and its opening point untangles the rest:

"Four separate things travel under the word install." AxonBuild, Install Claude Code When You Are Not a Developer

A terminal tool, a desktop app, an editor extension and a web version. A guide that picks one without saying so is how a project manager ends up following a developer's instructions for a job that never needed them.

The four questions, in order

If you answered no to question one, you are done with installing. The rest of this section is for the terminal route, and it is shorter than it looks.

Your windowPaste this line, then press Enter
Mac Terminalcurl -fsSL https://claude.ai/install.sh | bash
Windows PowerShellirm https://claude.ai/install.ps1 | iex
Windows CMDcurl -fsSL https://claude.ai/install.cmd -o install.cmd && install.cmd && del install.cmd

When it finishes, close the window, open a new one, and type claude --version. A working install prints a version number. That number is your proof, and it takes ten seconds to get. For a fuller check, claude doctor reports the health of the install without starting a session.

The two errors first-timers hit most are both about the window, not the tool. If you see 'irm' is not recognized, you are in CMD, so use the CMD line or reopen PowerShell. If you see The token '&&' is not a valid statement separator, it is the other way round. Anthropic's terminal guide for new users covers both, along with the other common first-day errors.

Last, type claude and press Enter. A browser window opens so you can sign in. That is a separate step on purpose: installing the tool does not log you in.

What has to exist before any of this is worth it

A folder of your own work. Claude Code starts in a folder you choose and works on the files there. In the terminal, that is whichever folder the window is open in. In the desktop app, it is the folder you pick with the Select folder button.

For a Scrum Master, that folder can be small. A week of plain standup notes is enough, like the five file week from our last post. Without a folder, you have installed a very capable tool with nothing to read, and the first session turns into a blank-page problem rather than a test of the tool.

It also helps to know what the install leaves alone. It does not change your existing files: in the terminal, Claude asks before it creates or changes a file, and in the desktop app's Manual mode, nothing is modified until you accept the change. It does not need administrator rights on Windows. And it does not sign you in. That is its own step, after the install.

"I am not opening a terminal. That is what IT is for."

I run ceremonies and manage stakeholders. If a tool needs me to type commands into a black window, it is not a tool for me.

Then do not open one. The desktop app is not a reduced version for beginners. Anthropic describes it as the same engine with a graphical interface. You choose the folder with a button, type requests in plain English, and in Manual mode you see each proposed change side by side and approve it before anything happens.

The choice is also less permanent than it feels. The desktop app and the terminal tool read the same configuration files, so the instructions, skills and settings you build in one carry over to the other. Starting in the app costs you nothing if you move to the terminal later.

"My laptop is locked down. I cannot install anything."

Corporate laptop, no admin rights. Every install goes through a ticket that takes three weeks.

The native installer does not need administrator rights. Anthropic's Windows instructions say so directly, and the tool installs into your own user folder rather than system folders. The requirements are modest: Windows 10 version 1809 or later, or macOS 13 or later, with 4 GB of memory and an internet connection.

So a missing admin password, on its own, does not stop you. Whether you should go ahead is a different question.

"It will break the first time it updates."

Every tool I have installed in the last two years has broken after an update, and nobody could tell me why.

That depends almost entirely on question four. The native install checks for updates when it starts and while it runs, downloads them in the background, and applies them the next time you open it. The Homebrew and WinGet versions never update unless you tell them to, which is how people end up months behind and following instructions for features they do not have.

If you would rather not be first, there is a setting that keeps you on a stable channel, typically about a week behind and skipping releases with major regressions. Either way, claude doctor shows whether the most recent update attempt worked, so a failed update is something you can see rather than guess at.

The part worth keeping

The install was never the hard part. The decisions were: which way in, which plan, which window, which update path. Answer those four and the command is a single line. Skip them and you get a developer's install for a job that did not need one, or a CMD error that looks like a broken tool when it is only the wrong window.

Once it runs, the first useful thing to do with it is small, and it says more about whether this tool is worth your time than any review will.

Sources: Anthropic, Claude Code advanced setup | Anthropic, terminal guide for new users | Anthropic, desktop app quickstart | Anthropic, desktop app reference | AxonBuild, Install Claude Code When You Are Not a Developer. All checked 2 October 2026.

AI Systems with Claude, for Scrum Masters and Project Managers

Most AI pilots in project management do not fail on the model. They fail because somebody pointed the thing at a decision instead of at a task. This course is built around that distinction, and around the work that follows once you get it right.

Seven sections, taught against one running project from the first lecture to the last. You watch a system get built, then you build the same one against your own sprint. Nothing here is a demo that works only on the example.

You finish with five working systems and you keep them. The sprint report machine turns your board export and standup notes into the report you currently write by hand. The retro intelligence system tells you what the team keeps saying, not just what it said this time. The stakeholder comms engine drafts the update in the register the audience expects. The backlog health monitor flags the quiet decay nobody has time to check for. The risk and dependency tracker follows the chains that actually bite.

Seventy-six working files ship with it, across four sections, so every system is built against real sprint data rather than material invented for a slide. Four sprints of retrospectives, board exports, standup notes, backlog and risk data.

Explore the Course


Build the Sprint Reporting System, Not Another Prompt

AI Systems with Claude is our own course for Scrum Masters and Agile Project Managers, taught against one running project across seven sections. You build five working systems and keep them: the sprint report machine, the retro intelligence system, the stakeholder comms engine, the backlog health monitor, and the risk and dependency tracker. Direct from HK School of Management, not on Udemy.

See What Is Included