What Claude Code Is & How It Thinks
Before you touch a keyboard, let's get the big idea — an AI that does things, not just talks.
- What "Claude Code" actually is, in plain words
- How it's different from a chatbot like the ChatGPT website
- The loop it follows: read → plan → act → ask permission
- Why it asks before changing your files (and why that's good)
The one-sentence version
Claude Code is an AI assistant that lives in your terminal and can read your files, write and edit code, and run commands to finish a task you describe in plain English. You stay in charge — it asks permission before doing anything that changes your computer.
The terminal (also called the command line) is a text window where you type instructions to your computer instead of clicking buttons. It looks intimidating, but you'll only need a handful of commands — and Claude Code does most of the typing for you. Module 2 walks you through opening it.
Chatbot vs. agent — the key difference
You may have used an AI chatbot on a website: you type a question, it types an answer, and that's the end of it. If it writes you some code, you have to copy it, paste it into the right file, save it, and run it yourself.
Claude Code is what we call an agent. The difference is that it can take the actions itself — open the file, make the change, run the test — and then tell you what it did. It's the difference between a recipe and a cook.
"Here's the code to add a login button. Copy this into your header.html file…"
You do the work: find the file, paste, save, test, fix the typo it caused.
"I'll add a login button to header.html. Here's the change I'm about to make — approve it?"
It does the work and shows you the exact change for approval before saving.
How it works, step by step
When you give Claude Code a task, it runs a simple loop. Understanding this loop is most of what you need to use it well.
-
It reads your project
Claude Code looks at the files in the folder you opened — your code, notes, and configuration — so it understands the context. It only sees what's in that project folder.
-
It makes a plan
It figures out the steps needed: which files to change, what command to run. For bigger tasks you can ask to see this plan before any action (that's "Plan mode," covered in Module 3).
-
It acts — and asks permission
When it wants to edit a file or run a command, it shows you exactly what it's about to do and waits for you to approve. Nothing changes on your computer without your "yes."
-
It checks its work and reports back
It runs tests or re-reads the result, tells you what happened, and waits for your next instruction. If something failed, it'll say so.
What's actually under the hood
Claude Code is powered by Claude, a large language model (LLM) made by Anthropic. An LLM is a program trained on enormous amounts of text that has become very good at predicting helpful, well-formed responses. It is not a search engine and it is not a person — it's a powerful pattern-completion engine wrapped in tools that let it act.
Large Language Model (LLM)
The AI "brain." Claude Code uses Anthropic's Claude models. You can pick a faster/cheaper model or a more capable one depending on the task (more in Module 2).
Agent
An AI that can use tools to take actions (read files, run commands) — not just generate text.
Context
Everything Claude Code currently "knows" in this session — your instructions plus the files and output it has looked at. It has a limited working memory, so context matters.
Permission / approval
Your "yes" before an action runs. The safety control that keeps you in charge.
Because it's a pattern engine, Claude Code can sometimes be confidently wrong — it may invent a function that doesn't exist or misremember a detail. This isn't a reason to avoid it; it's the reason you stay in the loop and review its work. We'll make that a habit in Module 5.
Just thinking exercise for now: picture a small task you do on a computer that involves following steps — renaming a batch of files, fixing a typo across a document, organizing a folder. That's the kind of thing you'll soon hand to Claude Code. Keep it in mind; you'll try a real one in Module 2.
Knowledge check
What's the main difference between Claude Code and a chatbot on a website?
Key takeaways
- Claude Code is an AI agent in your terminal that completes real tasks, not just a chat window.
- It follows a loop: read → plan → act → report, and asks permission before changing anything.
- It's powered by an LLM, which means it's powerful but can be confidently wrong — so you review its work.
- You are always the supervisor. Nothing happens without your "yes."
Finished this module?