CourseGuidesArticlesPodcastGitHubAboutStart free course
Course/What an AI Agent Is

7 min · +100 XP

Meet Claude Code

How it works on a task and where you can run it.

By the end of this lesson

You will be able to describe how Claude Code handles a task and choose where to run it.

Claude Code is Anthropic's agent for working on software and files. You describe a task in plain words. It reads your project, makes changes, runs commands and checks the result.

The loop it works in

Three phases, as the documentation describes them

  1. 1Gather context: it reads files and searches the project to understand what is there.
  2. 2Take action: it edits or creates files and runs commands.
  3. 3Verify results: it runs the program or tests, reads the output and decides what to do next.
  4. 4It repeats these phases until the task is done. You can interrupt at any point with Esc and steer it in another direction.
A short session
you ▸ make a page that shows today's date and a big button that counts clicks
claude ▸ The folder is empty, so I'll create a single index.html.
claude ▸ Write(index.html)
claude ▸ ⎿ Wrote 52 lines to index.html
claude ▸ Bash(open index.html)
claude ▸ Done. index.html shows today's date and a button with a counter.
claude ▸ Open it in your browser and click the button to check it counts.
For beginners

Real sessions show more text while Claude works. You do not need to read every line. Read its questions to you and the summary at the end.

Where it runs

Places you can use Claude Code

  1. 1Terminal: the command-line version with every feature. This course mainly uses it.
  2. 2Desktop app: the Code tab in the Claude app for Mac and Windows (Linux in beta). No terminal needed.
  3. 3VS Code and JetBrains: extensions that put Claude Code inside the code editor.
  4. 4Web: claude.ai/code runs sessions in the cloud and works with your GitHub repositories.
  5. 5Phone: the Code tab in the Claude app for iOS and Android starts cloud sessions or steers a session running on your computer.
ℹWhy this course uses the terminal

The terminal version has the full set of features, and the agent loop is the same everywhere. If the terminal feels like too much at first, the desktop app is an easier start, and most lessons still apply to it.

For experienced users
✦If you are experienced

It reads the codebase itself, so you do not paste files in; point it at a task and let it search. Git is conversational: "what did I change", "commit with a descriptive message", "resolve the merge conflicts". It runs your tests and builds to check its work, so tell it which command proves success. Sessions are saved locally: claude --continue picks up the last one in this folder.

Quick check

You ask Claude Code to fix a Save button that does nothing. What will it most likely do first?

Quick check

You do not want to use a terminal at all yet. Where can you start?

Facts checked against the official Claude Code docs October 24, 2026 · source

EAEvgenii Arsentev

Author

Evgenii Arsentev

PhD · Chief Executive Officer, digital health