Getting Started with Claude Code
Install Claude Code, run your first conversation, and understand the permission model. Everything you need to go from zero to productive.
What Claude Code is and why it matters
Claude Code is an agentic coding tool that lives in your terminal. Unlike chat-based AI assistants where you copy and paste code back and forth, Claude Code reads your files directly, makes edits in place, and runs commands on your behalf. It understands your entire project — not just the snippet you paste into a chat window. This means you can say "add error handling to the API routes" and Claude Code will find the relevant files, read them, understand the patterns, and make the changes. No copy-paste. No context switching. No explaining your folder structure every time. For non-developers, this is transformative. You can ask Claude Code to explain what a project does, reorganise files, update documentation, or automate repetitive tasks — all through plain English conversation. For developers, it is a force multiplier that handles the tedious parts of coding while you focus on architecture and design decisions. The key insight: Claude Code does not replace your thinking. It replaces your typing.
Installing Claude Code
Claude Code installs via npm, the Node.js package manager. If you do not have Node.js installed, download it from nodejs.org (the LTS version). Then open your terminal and run: npm install -g @anthropic-ai/claude-code. That is it. One command. To verify the installation, type: claude --version. You should see a version number. If you get a "command not found" error, your Node.js installation may not be on your system PATH — restart your terminal and try again. On macOS, you may need to use sudo: sudo npm install -g @anthropic-ai/claude-code. On Windows, use an administrator terminal. Claude Code works in any terminal: the built-in macOS Terminal, iTerm2, Windows Terminal, or the integrated terminal in VS Code. Once installed, navigate to any project directory and type: claude. This starts a new conversation with Claude Code in the context of that directory. Claude Code will automatically read your project files to understand what you are working with.
Your first conversation
Navigate to a project folder in your terminal and type: claude. Claude Code will start and you will see a prompt where you can type. Try asking: "What does this project do? Give me a summary." Claude Code will read your files — package.json, README, source files — and give you a structured overview. This is one of the most powerful first uses: understanding an unfamiliar codebase. Next, try: "What are the main dependencies and what does each one do?" or "Walk me through the folder structure." Claude Code reads files on demand, so it only looks at what it needs. You will see it request permission to read files — this is the permission model at work. For your first real interaction, try something low-stakes: "Add a comment at the top of each file explaining what it does." Watch how Claude Code reads each file, understands its purpose, and writes a contextual comment — not a generic one. This is fundamentally different from copy-paste AI coding. Claude Code has context.
Understanding the permission model
Claude Code operates on an explicit permission model. Before it reads a file, edits a file, or runs a command, it asks for your approval. You will see prompts like: "Claude wants to read src/index.ts — Allow / Deny." This is a safety feature, not a nuisance. It means Claude Code cannot silently modify files or run destructive commands without your knowledge. You have several options for managing permissions. You can approve individual actions one at a time. You can approve all actions of a certain type for the session (e.g., "allow all file reads"). You can also configure allowed and denied tools in your project settings for a more permanent setup. The permission model is especially important when Claude Code wants to run shell commands. A command like "npm install" is generally safe. A command like "rm -rf" requires scrutiny. The model trains you to be deliberate about what you allow — which is good practice even outside of AI tools. Think of it as a co-pilot that always asks before touching the controls.
Core commands
Claude Code has a small set of built-in slash commands that help you manage your session. /help shows all available commands and a quick reference for what Claude Code can do. /clear wipes the current conversation history and starts fresh — useful when the context gets cluttered or you want to switch tasks. /compact compresses the conversation to save context window space. When you are deep into a long session and Claude Code starts losing track of earlier instructions, /compact summarises everything and frees up room. /cost shows how many tokens you have used in the current session and what it has cost. This is valuable for understanding your usage patterns and managing API spend. /init creates a CLAUDE.md file for your project — we will cover this in the next guide. These commands are typed directly into the Claude Code prompt, just like chat messages. They execute instantly and do not count as conversation turns. Learn these five commands and you have the foundation for efficient Claude Code usage.
Your first real task
Now that you understand the basics, try a real task. Here are three good starter tasks that are low-risk and high-reward. First, file organisation: "Rename all the files in the utils/ folder to use kebab-case instead of camelCase." Claude Code will find the files, rename them, and update all imports across your project automatically. Second, documentation: "Create a README.md that explains what this project does, how to install it, and how to run it." Claude Code reads your code, your package.json, and your config files, then writes documentation that is actually accurate. Third, small automation: "Write a script that checks all links in our markdown files and reports any that are broken." Claude Code will write and test the script. The pattern for effective first tasks: choose something that is tedious for a human but straightforward for an AI with full project context. Avoid tasks that require deep architectural decisions — save those for after you have built trust in the tool. Start small, verify the output, and gradually increase scope.
Claude Code for Non-Developers
This guide is hands-on and practical. The full curriculum covers the conceptual foundations in depth with structured lessons and quizzes.
Go to lesson