Claude Code isn't ChatGPT with a terminal window wrapped around it. It's a different animal entirely — and once you understand the full landscape of what it can do, you leverage it completely differently.

This is a breakdown of every core concept from "what even IS this thing" all the way to skills, hooks, MCP, subagents, and scheduling. 30 concepts total. No filler.

Section 1 — Getting Started

1. The Terminal. Forget browsers. Forget GUIs. Claude Code lives in your terminal — the same place your shell scripts and system commands live. No buttons, no menus, no drag-and-drop. Just you typing what you want and Claude doing it across every file, folder, and app on your machine.

The tradeoff: one window instead of fifteen. You describe the outcome. Claude reads your files, edits them in place, shows you what changed. Done.

On Mac, open Launchpad, type "Terminal," hit Enter. Then mkdir ~/playground and cd into it. That's your lab.

2. Installation. One command.

Mac/Linux: curl -fsSL https://claude.ai/install.sh | bash
Windows:   irm https://claude.ai/install.ps1 | iex

Two pricing tiers matter: Pro at $20/month (limited) and Max at $100 or $200/month (unlimited). If you're using this for business daily, Max is the move. ROI shows up in week one.

3. File Access. This is where Claude Code stops being a chatbot and starts being useful. It reads and edits your actual files — not pasted copies of them. Your proposals, your spreadsheets, your client folders. It opens the file, reads it, rewrites the section inside your document. No copy-paste. No "please summarize what I just sent you." It SEES the file. And it creates files too — analyses, summaries, whatever you need.

4. Image + PDF Reading. Screenshots, scanned documents, whiteboard photos, receipts, diagrams. Drop it in, Claude reads it, works with it like text. No re-describing images to a chatbot and hoping it gets it right.

Section 2 — First Real Tasks

5. Tool Use. Claude doesn't chat. It ACTS. Reads files, edits documents, runs commands, searches folders, grabs web pages. Each action is a tool call you see happening in real time.

The key shift: AI stops giving advice and starts doing work. "Here's your updated spreadsheet, 47 rows changed" instead of "here are the steps to update your spreadsheet, good luck."

6. How to Talk to It. Vague prompts get vague results. "Help me with marketing" is garbage. "Write a 3-email welcome sequence for my dog walking business, casual tone, GPS tracking feature, under 200 words each, CTA to book a free walk" gets you results.

Specific ask, specific output. Use @ to tag files in your working directory — autocomplete shows the list.

7. CLAUDE.md. Your AI's operating instructions. One file, read at the start of every session. Business rules, brand voice, tools you use, things to never do. Write it once, Claude follows it forever. No more repeating yourself every session.

8. Plan Mode. Before Claude touches anything, it writes out its full approach. You review. You adjust. YOU approve. Then it builds. Shift+Tab to engage it.

Ninety percent planning, ten percent building. The reason your kitchen doesn't get demolished while you're getting a quote.

Section 3 — How the Brain Works

9. Context Window. Claude's memory is a whiteboard. Everything — every prompt, every file read, every response — goes on it. Big, but not infinite. When it fills up, older content gets summarized to make room.

Long rambling conversations fill it fast. Short, focused sessions keep it clean. Type /usage to check capacity.

10. Tokens + Cost. Every word costs tokens. Input (what you send) and output (what Claude sends back) both count. Four ways to manage cost:

  1. Use a cheaper model for simple tasks
  2. /clear between unrelated tasks so old context doesn't pile up
  3. /compact during long sessions to compress without losing key details
  4. Install the context-mode plugin

11. Model Selection. Three brains:

  • Opus — smartest, most expensive. Complex projects and big rewrites.
  • Sonnet — fast, capable, affordable. Your daily driver.
  • Haiku — cheap and quick. Simple questions and lookups.

Switch mid-conversation based on the task. Building a marketing strategy? Opus. Renaming a file? Haiku.

Section 4 — Managing Sessions

12. /compact. Summarizes the conversation into a condensed version. Frees up whiteboard space without losing the important stuff.

13. /clear. Wipes the current session entirely. Clean slate — use when switching to a completely different task so old context doesn't bleed into new work.

14. Session Persistence. Every conversation is a session. Close the terminal, come back tomorrow, run claude --resume and you're exactly where you left off.

Section 5 — Control

15. Permission Modes. From "ask before every edit" to "full autonomy." You set your comfort level and match your pace. New users keep tight guardrails. Experienced operators loosen them over time.

16. Effort Levels. Tell Claude how hard to think. Quick question gets a quick answer. Complex strategy gets deep analysis. There's a slider. And there's ultrathink — a keyword that maxes out reasoning for one response.

17. Escape to Interrupt. Press Escape anytime. Claude stops on a dime. No work lost. No runaway processes. You redirect and keep moving.

Section 6 — Review + Teaching

18. VS Code. For reviewing Claude's work across multiple files, VS Code is the upgrade from the terminal. Color-coded diffs, side-by-side comparisons, sidebar navigation.

Terminal for giving instructions; VS Code for reviewing results.

19. Memory. CLAUDE.md is your instruction manual — system-level rules and workflows. Memory is Claude's personal notebook — preferences, corrections, context about ongoing work.

Correct it once, Claude writes it down. It never forgets. System stuff goes in CLAUDE.md. Preferences accumulate in Memory.

20. Scope. Project scope is per-project settings and rules (each folder gets its own CLAUDE.md). Global scope is settings that follow you everywhere — writing style, name, preferences.

Section 7 — Skills + Automation

21. Slash Commands. Quick actions you'll use constantly: /help, /clear, /compact, /model, /usage).

Hidden gems: /insights (usage patterns and a personal performance dashboard) and /btw (side-note without interrupting Claude's current task).

22. Skills. This is where Claude Code goes from useful to life-changing.

A skill is a saved workflow — instructions as a markdown file that triggers an entire chain of actions. One command, dozens of steps. Search videos, transcribe audio, write captions, run quality checks, publish to multiple platforms, update databases.

Build once. Run forever. Improve over time.

23. Hooks. Automated guardrails that run before or after actions. Quality gates that block content with banned words, exceeded character limits, missing media. Desktop notifications when long tasks finish.

Set the rules once, hooks enforce them every time.

Section 8 — Real World Connection

24. Web Browsing. Paste a URL, Claude reads the page, extracts data, takes action. Not just summarizing — writing comparison docs, saving analyses, moving to the next task. One URL becomes a finished document.

25. MCP Servers. Model Context Protocol — the bridge between Claude and your real tools. Google Drive, Slack, Notion, Airtable, Stripe, email.

Without MCP: "here's what you should update in your spreadsheet." With MCP: "your spreadsheet is updated, here's what changed."

26. Perplexity MCP. Read one page, or research the entire internet with verified sources. Perplexity MCP finds the best sources, synthesizes them, cites every claim. From reading one link to having a research assistant who reads fifty and gives you the summary with citations.

Section 9 — Agents, Remote, Scheduling

27. Subagents. Give Claude five independent tasks and it spawns five mini-Claudes to handle them simultaneously. Sequential becomes parallel automatically. No setup required — Claude decides when parallel work is faster.

28. Remote Control. Start a session on your computer, continue from your mobile phone. Walk away from your desk, keep working from the couch or a coffee shop. The computer does the heavy lifting; the phone is the remote.

Run claude remote-control, scan the QR code with your phone camera.

29. /loop (Scheduling). Schedule recurring Claude tasks at intervals you set. Terminal sessions only — close the terminal and they stop. For persistent scheduling, use a dedicated task runner.

30. Git Version Control. Built-in. Every change is a save point. Roll back to any previous version in seconds. Local git covers solo work. Add GitHub for cloud backup and team collaboration.

Where to Start

Week one: Install it. Watch it read and edit a file. Watch the tools work. Build a CLAUDE.md with three rules. Use plan mode for your first real job.

Week two: Learn context management and cost control. Switch models. Set up memory. Build your first simple skill. Try VS Code for review.

Month one: Connect an MCP server. Add Perplexity MCP. Install the context-mode plugin. Set up remote control. Schedule a recurring task. Check /insights to optimize.

Prebuilt skills are already available on GitHub — official Anthropic repos, community directories, 180+ skills for engineering, marketing, compliance. Drop the URL into Claude Code and have it set up the three most relevant ones.

Start. Open the terminal. Type claude. Build something.

Inspired by Every Claude Code Concept Explained for Normal People by Sabrina Ramonov. This is our rewrite — same concepts, different words.