The life agentic

Claude Code

claude is Anthropic’s agentic coding tool for the terminal.

For students who want one strong paid option, Claude Code with a Claude Pro subscription is the clearest place to start. It is a good fit when GitHub student access is unavailable, when you want to start immediately, or when you want a polished default workflow for repository work.

If you are comparing Claude Code with GitHub Copilot, direct APIs, OpenCode Zen, or local models, see Model access.

Intended learning outcomes covered on this page

After working through this page, students should be better able to:

• choose an appropriate tool for a task by distinguishing between coding-agent workflows and scripting or pipeline workflows
• explain why repository-grounded terminal agents support different kinds of work than ordinary browser chat
• use a coding agent in a way that separates planning from editing
• recognise context-window limits and use strategies such as fresh sessions or compaction when context quality degrades
• explain how subagents or separate runs can isolate side work and keep a main task focused
• distinguish between persistent instructions and reusable memory or context
• use AGENTS.md and skills appropriately as reusable instruction layers in a project workflow
• critically verify model output against files, tools, and project context before relying on it

What it is good for

• exploring an unfamiliar repository
• planning a change before editing files
• making multi-file changes with approval prompts
• reviewing diffs, commits, and pull requests
• working with reusable instructions through CLAUDE.md and skills

Before you install it

You should already have:

1. a terminal you are comfortable using
2. a repository or project directory you trust
3. an account that includes Claude Code access

For most students paying themselves, Claude Pro is the simplest starting point. Claude Code also works with Claude Max, team plans, Console accounts, and some cloud-provider setups, but those are not usually the first thing to learn.

Installation

Claude documents several installation methods. Common choices are:

Native install on macOS, Linux, or WSL

curl -fsSL https://claude.ai/install.sh | bash

Homebrew on macOS or Linux

brew install --cask claude-code

Windows PowerShell

irm https://claude.ai/install.ps1 | iex

For native Windows setup, Anthropic currently requires Git for Windows to be installed first.

First start

Start Claude Code inside a repository or project directory:

claude

On first launch it will usually ask you to log in in the browser. After that, you can reopen it with claude in any project you trust.

A safe first workflow

A good beginner workflow is:

1. ask Claude Code to explain the project before you ask it to edit anything
2. ask for a plan before making a non-trivial change
3. approve file edits and shell commands carefully
4. inspect the resulting diff yourself

Good first prompts are:

Explain the structure of this project.

Find the main entry point and explain how the program starts.

Plan a small improvement to the README, but do not change any files yet.

Approvals and session hygiene

Claude Code is much more useful when you treat a session as one task at a time.

Good habits:

• start with explanation or planning prompts
• read permission requests before approving them
• use version control so you can inspect what changed
• keep one task per session when possible
• if a session gets long or messy, use /compact or exit and start a fresh one
• treat the output as capable but fallible

For a more explicit student workflow for checking repository claims, diffs, commands, and tests, read Verification.

CLAUDE.md, skills, and AGENTS.md

Three related ideas matter quickly:

• CLAUDE.md gives Claude Code persistent instructions that load every session
• skills give Claude reusable playbooks that load only when relevant or when you invoke them
• AGENTS.md is a more general cross-agent instructions file used by many tools

Claude Code reads CLAUDE.md, not AGENTS.md, directly. If a project already uses AGENTS.md, the usual bridge is to create a CLAUDE.md that imports it.

See also:

• Skills
• AGENTS.md

Notice the pattern

If you compare Claude Code with the other coding-agent tools in this guide, the same recurring ideas show up again:

• Planning before editing: Claude Code works well when you ask for a plan first and keep non-trivial changes read-only until the plan is sound.
• Limited context: one task per session, /compact, and fresh sessions all help manage a limited context window.
• Work isolation: subagents with their own context windows let side work stay out of the main thread.
• Instructions and memory: CLAUDE.md, skills, and AGENTS.md are instruction layers, while Claude’s project memory is separate.

After you have seen the three main coding-agent guides, the page on Agentic concepts reconnects these recurring ideas.

Short version

1. Get a Claude plan that includes Claude Code access.
2. Install claude.
3. Start it inside a repository you trust.
4. Begin with explanation and planning tasks.
5. Approve write actions carefully.
6. Learn CLAUDE.md, skills, and AGENTS.md once the basics feel natural.

Next step

• Understand model access: GitHub, Claude, APIs, Zen, and Ollama
• Learn what skills are
• Learn what AGENTS.md is for
• Learn a reusable verification workflow
• Get started with OpenCode
• Understand the shared agentic concepts
• Which tool should I use?

Official links

• Claude Code quickstart
• Claude Code setup
• How Claude remembers your project
• Claude Code skills

This site is open source. Improve this page.