Claude Code works well for students who set it up to teach rather than to do the work for them: switch on the Learning output style, start bigger tasks in plan mode, and ask it to explain code before it writes any. It needs a paid Claude plan, because the Free plan doesn't include it. If your university has Anthropic's Education plan, ask whether it covers Claude Code.
Claude Code is Anthropic's coding agent. It reads your project, edits files and runs commands in your terminal, the desktop app, VS Code or the browser. That makes it far more capable than pasting code into a chat, and also easier to lean on too hard. This guide covers the settings that keep you learning, which plan you need, how to use it safely on your own laptop, and prompts that build skill.
How do you turn on learning mode in Claude Code?
Use an output style. Anthropic's output styles docs describe four built-in styles besides Default. Two are made for learning:
| Style | What changes | Good for |
|---|---|---|
| Explanatory | Claude adds short Insight blocks that explain the choices behind the code it writes |
Understanding a new codebase or language |
| Learning | Same Insight blocks, plus Claude leaves small pieces of code for you to write | Hands-on practice while the task still gets done |
To switch, run this inside a session:
/output-style learning
You can also run /config and pick Output style, or set it in a settings file. The value is case-sensitive there, so write "outputStyle": "Learning". To make it your default in every project, put that line in ~/.claude/settings.json.
In the Learning style, Claude handles routine work itself. When it reaches a piece with a real design decision, such as error handling, a data structure or business logic with more than one valid approach, it adds a TODO(human) comment in the file and sends a Learn by Doing request that says what's built, what to write and what to weigh. Then it waits. Write your code at the marker, tell Claude you're done, and it responds with an Insight about your code before it continues.
The docs add a fair warning: an output style "gives Claude instructions to follow. It doesn't guarantee that something always happens or never happens." Treat it as a strong nudge, and ask for explanations when you want more.
Which Claude plan do students need?
Claude's pricing page says "Claude Code is included in all paid plans," and its feature table lists Claude Code as not included in Free. As of October 2026:
| Plan | Price (as of October 2026) | Claude Code |
|---|---|---|
| Free | $0 | Not included |
| Pro | $17 per month billed annually ($200 up front), or $20 billed monthly | Included |
| Max | From $100 per month, with 5x or 20x Pro usage | Included |
| Education | University-wide plan bought by your institution | Ask your school |
Usage limits are shared: the pricing FAQ says Claude Code "shares the same usage limits as the rest of your plan," so chat and terminal work draw from one pool. Run /usage in a session to see where you stand.
The Education plan is sold to institutions, not individuals. The pricing page describes it as "a comprehensive university-wide plan for an institution, including its students, faculty, and staff," with "Academic research and learning mode." What each university's contract includes varies, so ask your IT department or library whether Claude Code is part of it before you pay for Pro. Prices change; check the Claude Code pricing guide and the official page before you buy.
How should students use Claude Code without skipping the learning?
A workflow that keeps you in charge:
- Ask before you build. Anthropic's common workflows page suggests starting in a new project with "give me an overview of this codebase," then asking about its architecture patterns and key data models. This works just as well for a starter repo from your course.
- Plan first. Run
/planor press Shift+Tab until plan mode is on. Claude reads the code and proposes an approach without editing your files, so you can question it before anything changes. See the plan mode guide. - Write the hard parts. Keep the Learning style on and do the
TODO(human)pieces yourself. - Explain it back. After a change, ask Claude to quiz you on it, or explain it in your own words and ask Claude to correct you.
- Check your course rules. Policies on AI help differ between courses and assignments. When in doubt, ask your instructor and keep a record of how you used the tool.
New to the tool itself? Run /powerup, which the commands reference describes as quick interactive lessons with animated demos of Claude Code features.
How do you use Claude Code safely on your own laptop?
Claude Code runs commands with your user account, so a few defaults are worth keeping:
- Switch to Manual mode while you learn. In Anthropic's permission modes table, Manual mode (config value
default) lets Claude run only reads without asking, so you approve each edit and command and see what it does. Recent versions start terminal and VS Code sessions in auto mode, where a classifier reviews actions instead of you; press Shift+Tab once to switch to Manual. - Skip
bypassPermissions. Anthropic lists it as for "Isolated containers and VMs only." - Use checkpoints, and git too. Every prompt creates a checkpoint, and
/rewindor pressing Escape twice restores code, conversation or both. The docs note that a file changed only by Bash commands isn't tracked, and that checkpoints are "Not a replacement for version control." Commit often. The checkpoints guide has details. - Keep secrets out. Don't paste API keys or passwords into prompts, and keep
.envfiles out of your repo. - Clear between tasks.
/clearstarts fresh when you switch to unrelated work, which also saves usage.
Which prompts help you learn?
These are example prompts written for this guide, not taken from Anthropic's docs. Each one asks Claude to teach before it acts.
Explain how this project is structured: what each folder does, where the program starts, and how a request flows through it. Don't change any files.
My code in sort.py fails on the second test. Don't fix it. Ask me questions that help me find the bug myself, one at a time.
I wrote this function. Review it like a strict TA: correctness, edge cases, naming and complexity. Give me hints, not a rewritten version.
I'm learning recursion. Give me three small exercises in this repo's language, from easy to hard, and check my answers when I paste them.
The code explanation prompts, debugging prompts and teacher prompts collections have more ideas.
Where else can students learn Claude Code?
- Claude Academy. Anthropic's course list includes Claude Code 101 (12 lessons and a quiz, about 1.5 hours) and Claude Code in action.
- Campus programs. Anthropic runs Claude campus programs with student ambassadors who receive a USD3,600 stipend, subject to program terms. The page says the program is closed for Fall 2026.
- Other agents. If you also use OpenAI's tools, the Codex for beginners guide covers the same ground for Codex.
How do you keep the prompts that work for you?
The prompts that help you learn are worth saving. Kipt keeps prompts and Agent Skills as Markdown in a private GitHub repo you own, with {{name}} blanks for things like a file or topic, and installs your skills into Claude Code or Codex with npx skills add. To turn a study prompt into a reusable skill, try the free SKILL.md generator.