A coding partner that can actually touch the code
Most people meet AI through a chat box. You paste in a snippet, you get a snippet back, and you do the copying, running and fixing. Claude Code flips that arrangement. It is an agentic coding tool: it reads your codebase, edits files, runs commands and checks its own work, all inside your real project.
The key word is agentic. A chatbot answers. An agent acts, looks at what happened, and decides what to do next. Ask Claude Code to "fix the failing login test" and it will search for the test, read the code under test, run the test suite, edit the bug, rerun the suite and report back. You can watch every step and stop it at any time.
Chat assistant vs. coding agent
| Chat assistant | Claude Code | |
|---|---|---|
| Sees your project | Only what you paste | Searches and reads files itself |
| Makes changes | Suggests code for you to copy | Edits files directly (with your permission) |
| Runs things | No | Runs tests, builds, linters, git, CLIs |
| Checks its work | You do | It reruns the check and iterates |
| Remembers your conventions | Per chat | CLAUDE.md, skills and settings in the repo |
Where Claude Code runs
Claude Code is one engine with several front doors. Pick whichever fits the moment. Your project files, CLAUDE.md and settings work the same way everywhere.
- Terminal (CLI): the original and most powerful surface. Type
claudein any project folder. - VS Code and JetBrains: inline diffs,
@-mentions of files and plan review, right inside your editor. - Desktop app: parallel sessions with git isolation, diff review and previews.
- Web and mobile: run tasks in the cloud against a GitHub repo, then review the pull request, even from your phone.
- Integrations: Slack, GitHub Actions, GitLab CI/CD and a Chrome extension for testing web apps.
What Claude can't see (yet)
Claude Code is powerful, but it isn't psychic. It starts every session knowing nothing about your project beyond what it can read. It can't see your Jira ticket, the Slack thread where the team chose Postgres, or the reason that odd-looking function exists. A big part of this course is learning to tell Claude what it can't see and give it a way to check its own work. Those two habits separate frustrating sessions from magical ones.
A Claude subscription (Pro, Max, Team or Enterprise) or an Anthropic Console account with API credits, plus basic comfort with a terminal and a code editor. That's it. You don't need to be a senior engineer; Claude Code is a fantastic way to become one.
Check your understanding
What makes Claude Code agentic rather than a regular chat assistant?
Which of these can Claude Code NOT know unless you tell it?
Where can you run Claude Code?
Who approves actions that change things in your project?