# Pharos is taking you through this one.
# Free, and it runs right here. No signup.
$ learn Claude Code
Learn Claude Code in 5 minutes.
Claude Code is a coding agent that runs in your terminal, reading your repo, editing files, and running commands with your permission. You will practice writing task briefs, steering it with a CLAUDE.md file, and using plan mode and permissions to keep the leash exactly as long as you want.
# 5 minutes, no install
# interactive, not a video
# Lesson not loading? Open it in a new tab. It is the same free lesson either way.
How long is the leash?
An agent that can edit files and run commands is exactly as useful as it is trusted. Claude Code makes trust adjustable, and dialing it correctly per task is the core skill.
Plan mode
In plan mode, Claude Code investigates the codebase and proposes an approach without changing anything. For unfamiliar repos or risky work, you review the plan first and let execution start only once you agree with it.
Permission prompts
Every file edit and shell command is requested before it runs, and you can allow once, always, or never per tool and pattern. The prompts are the leash: generous for a scratch repo, tight for production code.
Standing instructions
A CLAUDE.md file at the repo root loads into every session: build commands, conventions, things to never touch. It is how the agent stops asking the same questions and stops repeating the same mistakes.
What you will do in the lesson
You answer and try things. Nobody demonstrates at you for five minutes.
- Prompt like a delegator. You rewrite vague requests into task briefs with success criteria, the difference between an agent that wanders and one that finishes.
- Run plan mode first. You review a proposed plan for a nontrivial change and catch the step that would have gone wrong.
- Set the permissions. You decide which commands deserve always-allow and which should prompt forever, matched to a real repo's risk profile.
- Write a CLAUDE.md. You draft standing instructions for a sample project and see how they change the agent's behavior on the next task.
- Review like it is a PR. You inspect an agent-authored diff, request a fix, and practice the judgment call of when to take the keyboard back.
Before you start
Why a terminal instead of an editor?
The terminal is where builds, tests, git, and deploys already live, and Claude Code drives all of them. It pairs with whatever editor you keep rather than replacing it, and it scripts and pipes like any other command-line tool.
Can it break my codebase?
Not without asking. Edits and commands go through permission prompts, work lands on your git branch where diffs are reviewable, and plan mode lets you approve intent before anything changes. The lesson treats configuring that safety net as a first-class skill, not a footnote.
Does this lesson require installing Claude Code?
No install and no API key: the lesson is interactive and runs right here in the page. If you set up the real thing later, it is a single command in your terminal.
I already have Copilot in my editor. What does this add?
Copilot assists while you type; Claude Code works while you do something else. You hand it a task, a bug, a refactor, a test suite, and review the result, terminal-native and multi-file by default. Developers who use both tend to split by size: keystrokes to Copilot, whole tasks to Claude Code.
Other tools people learn next
Same format, same five minutes, still free.
Learners from these organizations are already learning on 
