Claude Code overview
Claude Code is Anthropic's agentic coding tool: it reads your project, edits files, runs commands and checks its own work from your terminal.
Last updated: 29 Sep, 2026 · Claude Code
This course teaches it on one small repository you can read in full.
Anthropic released Claude Code as a research preview in February 2025 and made it generally available in May 2025. It runs in a terminal on macOS, Linux and Windows, and the same agent is available in VS Code and JetBrains IDEs, a desktop app and on the web. This course uses version 2.1 in the terminal.
Asking Claude Code to explain an existing project
The video starts where most developers start: a project somebody else already built. It opens an existing repository of LangChain notebooks in Cursor, types claude in the terminal, says yes when asked whether to trust the folder, and asks it to analyse the entire project. Claude Code works through every folder and file on its own and comes back with a summary: a bootcamp companion repository for LangChain 1.x, six tutorial notebooks from the intro to middleware plus a test script, with nothing committed yet.
The video calls this getting a knowledge transfer from the model instead of booking an hour with a teammate. That is the everyday value of the tool, and the rest of this course is about how it does it and how you control it.
The clips come from the Complete Claude Code Course In 2 Hours For Developers, recorded on Claude Code 2.1.150. Where a later release changed something a clip shows, a note under the clip says what changed.
Agent loop, CLAUDE.md, hooks and skills
It works as an agent loop. You describe a task; Claude reads files, searches the code, runs commands and edits files, one tool call at a time, and asks permission before actions you have not allowed. Each project can shape that loop with a CLAUDE.md file of instructions, settings with allow and deny rules, hooks that run your own scripts at fixed points, skills, subagents and MCP servers.
Using it needs a paid Claude plan or an Anthropic API key. It can also run without a terminal session, in headless mode for scripts and in GitHub Actions for review and automation, which later lessons cover.
Asking Claude Code a question is quick to learn. Leaving it alone with real work takes more: control over what it may do, what it remembers, which procedures it repeats, and how you find out what it did. Most of this course covers those parts.
The project you will work on
The course uses one tiny repository from the first lesson to the last: a link shortener that turns a long URL into a short code and back. It has three short files, no framework, and one bug that has gone unnoticed. You set it up in the next lesson. From there Claude Code searches it, writes tests for it, fixes the bug in an unattended run, and you fix it yourself in the last lesson.
CLAUDE.md, a command, a skill, a hook and CI review
By the last lesson the repository carries all of the following, each added deliberately:
| What | Why it is there |
|---|---|
A CLAUDE.md | So the rules of this project do not have to be repeated every session |
| A slash command | One line instead of a paragraph you keep retyping |
| A skill | A procedure Claude follows the same way every time |
| A hook | A rule that is code, not a request, so it cannot be talked out of |
| A subagent | Work that happens without filling up your conversation |
| A review in CI | Somebody reading the change when you are not |
Terminal captures and the browser panel
Each lesson adds one idea. Terminal blocks were captured by running the command against this repository. Where something happens inside the interactive interface and cannot be captured as text, the lesson describes it instead of showing a screenshot.
Related
- Next: Installation and setup
- Reference: Claude Code docs
- Install Claude Code (the next lesson) and ask it one question about your own machine.
- Think of a small repository you know well; that is the kind of project this course works on.
Every expert started right here.