Scope and Getting Started

What is this documentation for? Where does it stop? Both questions matter more than they might seem, because the answer decides whether you should read this whole page or skip straight to the workflow chapters instead.

Scope

This documentation covers installing and signing in to Claude Code, the terminal and graphical interfaces, core concepts like tokens and context, the CLAUDE.md configuration file, three common design approaches, six permission modes, Plan Mode, and scaling work across multiple agents. All of it comes from my own course walkthrough.

FIGURE 1 — Course topic map. Ten chapters in three reading stages, start anywhere a group answers your question.

It does not cover Agent Teams, Git Worktrees, or Session Mobility. I name these once, only in passing, and I explicitly defer a fuller explanation to a later part of the course that this transcript doesn't include. It doesn't cover backend architecture or team-scale tooling either. MCP (Model Context Protocol) is covered, in its own chapter, through three worked connections rather than the full protocol specification. I place the topics I do skip in modules that come later. This is not a general Claude Code reference. It reflects my own workflow, not Anthropic's own documentation, and the two will disagree here and there.

Getting set up

Installation runs through install commands specific to your operating system: macOS, Linux, WSL (Windows Subsystem for Linux), or Windows. Run them from a terminal. A first install on a brand-new machine can take a little longer than a reinstall does on a machine that already has the tooling in place.

Why local, not a browser tab? A local agent can read your project files, edit them, and run scripts as part of finishing a task, something no browser-based chat window can do, no matter how capable the model behind it happens to be.

You'll sign in one of three ways. Figure 2 lays out the choice: a Claude subscription (Pro, Max, Team, or Enterprise), an Anthropic Console account billed by API (application programming interface) usage, or a named third-party platform such as Amazon Bedrock. Pick based on how you already pay for tools. Nothing here suggests one route works better than the others technically. The difference is entirely about billing.

FIGURE 2 — Three ways to sign in to Claude Code. Pick the column that matches how you already pay for tools, since I found no technical reason to prefer one route over another.

Once you're in, the terminal shows a version number, the active model, your working directory, and a status line with token counts. Skip memorizing the exact version shown in my own demo. It's outdated already. Learn the shape of the interface instead, because that part doesn't change every release: a permission-mode indicator you cycle with Shift+Tab, and a processing message that rotates through different phrases while a request runs quietly in the background.

Two ideas come up constantly from here on, so get them straight now. A token isn't quite a word. Treat the two as roughly interchangeable while you're starting out, but know that punctuation and word-fragments can each cost a token of their own, which is why a short sentence sometimes costs more than its word count would suggest. Context is simpler: it's how much conversation history currently fills the model's available window, shown as a percent running from zero to a hundred. Claude Code compresses older history now and then, keeping that percentage from just climbing forever.

Choosing where you'll work

An IDE (integrated development environment) is three things bolted together: a file browser, a text editor, and an AI (artificial intelligence) chat panel, and nothing more technical hides behind that definition no matter how the term gets thrown around elsewhere. Three parts, no more.

FIGURE 3 — IDE = three parts. A file browser, a text editor, and an AI chat panel, nothing more technical hides behind the term.

I demonstrate two IDEs in the course, compared directly in Figure 4: Visual Studio Code, the established choice, Microsoft-built and heavily extensible with a large support community behind it, and Antigravity, newer and built on top of VS Code but leaning harder into AI-native workflows, with some features still rough at recording time. That comparison is a snapshot, not a verdict. Both tools will have moved on by the time you read this. My early chapters use VS Code; nearly everything after switches to Antigravity.

FIGURE 4 — Visual Studio Code vs. Antigravity, as of this recording. Treat this as a snapshot, not a verdict, since both tools will look different by the time you read this.

Claude Code doesn't ship inside either editor. You install it as an extension. Pause on this step. Verify it's the one published by Anthropic, with the verification checkmark, before you click install, because unofficial copies of popular extensions are a genuine malware vector, in this ecosystem and in every other one that supports plugins. Keep your first sessions simple too: use the graphical panel rather than juggling an editor, a terminal, and a chat window all at once. Terminal fluency can wait.