How to vibe code with Claude Code: a beginner's guide

How To Use Claude Code For BeginnersWatch

Claude Code is an AI coding agent from Anthropic that runs in your terminal: you type what you want in plain sentences, and it reads, writes, and edits the files in your project. It's the fastest-growing tool in vibe coding, and the terminal part is far less scary than it sounds: here's how to go from zero to shipping with it.

First, about the terminal

The terminal is just a text window that talks to your computer. On a Mac, press Cmd+Space, type “Terminal,” and press Enter. That black window is the whole mystery. You will type sentences into it exactly like a chat box. The difference from a chat website is that this chat can actually touch your files, which is the entire point.

Install and start

Anthropic's install is one command in the terminal (their docs keep the current one. Search “Claude Code install”). Once installed, the working pattern is: go to your project's folder, start the agent, talk.

cd my-project     # go into your project folder
claude            # start Claude Code

No project yet? Make an empty folder and tell Claude Code what you want to exist: “Create a website for my photography portfolio: home, gallery, contact form.” It scaffolds the whole thing and tells you what it did.

The workflow that works

Claude Code's superpower is multi-step tasks. Where editor-based tools shine at “change this thing I'm looking at,” Claude Code shines at “do this whole job”: build the feature, run it, notice the error, fix the error, run it again. So give it whole jobs, not fragments, and give it acceptance criteria, because an agent that knows what “done” means checks its own work.

Add a contact form to the site. Done means: it emails me
at hello@example.com, shows a thank-you message after
sending, and rejects empty submissions with a clear error.

When it finishes, it summarizes what changed. Read the summary every time. You don't need to read the code. You need to be able to restate what happened in one sentence. If you can't, say “explain that in plain language” before moving on.

The two habits that separate good from lucky

• Commit checkpoints. After anything works, say “commit this.” A commit is a save point: when a future change breaks things, you restore instead of untangling. Claude Code handles the git commands; you just have to ask.
• Write decisions down. Claude Code reads project instruction files (like AGENTS.md or CLAUDE.md) before working. Rules you put there (“never touch the payments code without asking,” “the design system lives in styles/”) get followed in every future session instead of living in your head.

Claude Code vs Cursor: which one?

Try Claude Code if you like delegating whole tasks and trust yourself to review summaries. Try Cursor if you want to see every change land in front of you as it happens: the visual feedback loop is better for building intuition. Plenty of builders use both: Cursor for hands-on sessions, Claude Code for “go do this while I make coffee.” The guide for the other side: How to vibe code with Cursor.