Course contents
What Is Claude Code
Claude Code is an agent that reads your project, edits files, runs commands and checks its own work. Learn what it can reach, where it runs, and what your job becomes.
Start with a prediction. You open Claude Code inside a project and type one line, “fix the failing tests,” with nothing else. Before reading on, decide what you think it does first.
You type “fix the failing tests” and nothing else. What does Claude Code most likely do first?
An agent, not autocomplete
Autocomplete predicts the next few characters in the file you have open. Claude Code works on the whole task. Anthropic's docs describe its work as three phases that blend together: gather context, take action and verify results. A question about your codebase may only need the first phase. A bug fix cycles through all three several times: run the tests, read the error, search for the code, edit it, then run the tests again.
Two parts make that loop work. The model does the reasoning; as of November 2026 the default on every plan is Claude Opus 5.5. The tools do the acting: reading and editing files, running shell commands, searching the web. Claude Code is the layer around the model that provides those tools, decides what the model sees, and enforces your permission settings. The docs call that layer the agentic harness, and it is the same harness on every surface you'll use.
| Tool category | What Claude can do |
|---|---|
| File operations | Read files, edit code, create files, rename and reorganize |
| Search | Find files by name, search their contents, explore a codebase |
| Execution | Run shell commands, start servers, run tests, use git |
| Web | Search the web, fetch documentation, look up error messages |
| Code intelligence | See type errors after edits, jump to definitions, find references (needs a code intelligence plugin) |
Match each request to the kind of tool Claude reaches for.
Pick one on the left, then its partner on the right.
What it can reach
When you run claude in a folder, Claude Code can work with everything below. That reach is the point: it is why Claude can fix a bug that spans five files when an editor plugin sees only the one you have open.
- Your project: the folder you launched from and its subfolders, plus other folders if you grant them.
- Your terminal: any command you could run yourself, such as build tools, git and package managers.
- Your git state: the current branch, uncommitted changes and recent commits.
- Your instructions: a
CLAUDE.mdfile, or anAGENTS.mdwhen the project has noCLAUDE.md. - Auto memory: notes Claude saved for itself in earlier sessions, such as preferences you corrected.
- Extensions you add: MCP servers, skills, subagents, hooks and plugins, each covered later in this course.
Esc twice (or running /rewind) can undo the change. Permission modes decide what runs without asking. Since v2.1.283, new terminal and VS Code sessions start in auto mode, where a separate classifier model reviews actions and blocks risky ones such as curl | bash or a force push. Later lessons show you how to tighten or loosen both.One engine, many surfaces
Every surface connects to the same Claude Code engine, so your CLAUDE.md, settings and MCP servers work across all of them. What changes is where the code runs. A local session runs on your machine. A cloud session runs in a virtual machine Anthropic manages and keeps going with your laptop closed. Remote Control lets you drive a local session from a phone or browser while the commands still run on your machine.
| Surface | What it's good for |
|---|---|
| Terminal (the CLI) | The full-featured original. Everything in this course works here first. |
| VS Code and JetBrains | Inline diffs and @-mentions inside your editor. The JetBrains plugin runs the CLI in the IDE terminal. |
| Desktop app (the Code tab) | Several sessions side by side, visual diff review and app previews. Includes Claude Code, so no separate install. |
| claude.ai/code and the Claude mobile app | Cloud sessions you start, check and steer from a browser or phone. |
| Slack | Mention @Claude with a bug report in a channel and get a pull request back. |
| GitHub Actions and GitLab CI/CD | Mention @claude in an issue or pull request, or run automation on repository events. |
Where do the commands actually run? Sort each way of using Claude Code.
A session you start at claude.ai/code
What your job becomes
Vibe coding is working at the level of outcomes: you describe what should be true, and the agent handles the files, commands and searches. It doesn't mean giving up responsibility. Your job moves to four things the agent can't do for you: say what you want precisely, define what “done” means, read the evidence, and decide what ships.
Anthropic's best-practices guide puts the most useful habit plainly. Claude stops when the work looks done. If it has something to run that comes back pass or fail, such as a test file, a build or a linter, it can check its own work and keep going until the check passes. Without one, you become the test suite.
Which single thing most improves Claude's chances of finishing a task correctly while you aren't watching?
The team that builds Claude Code explains the agentic model this whole course is built on — worth 10 minutes before you type your first prompt.
Plain-English definitions from the AI Coding Dictionary.