How to Use Claude Code: A Guide From a Real Project

How do you use Claude Code? In short: open a terminal in your project folder, type claude, and drive the rest through conversation. Claude Code is Anthropic's AI coding tool that reads your codebase, edits files and runs commands. We wrote this guide from real experience, building this website and its internal portal with it.

In short:

• Installation is one command; then you run claude in your project folder and log in.
• A CLAUDE.md file reminds Claude of your project's rules in every session; /init writes the first version.
• Permission modes decide what runs without asking; switch them with Shift+Tab in the terminal.
• The best results come from making it verify its work: build, tests and a preview.

What is Claude Code?

Claude Code is Anthropic's agentic coding tool. It reads your codebase, edits files, runs commands and works with your development tools (Claude Code docs). You can use it in the terminal, in the VS Code and JetBrains extensions, in the desktop app and in the browser.

Unlike classic autocomplete, it can take on a whole task instead of suggesting one line. Ask it to find and fix a bug, and it locates the relevant files, makes the change and runs the tests. You see every step and approve or redirect as needed.

How do you install Claude Code?

Installation takes one command. On macOS, Linux and WSL:

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

On Windows PowerShell:

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

Then go to your project folder and type claude. You are asked to log in on first use. Most surfaces require a Claude subscription or an Anthropic Console account.

How do you teach Claude Code about your project?

Every session starts with a fresh context. To avoid explaining your project's rules each time, you use a CLAUDE.md file. It holds build and test commands, coding conventions and architectural decisions, and Claude Code reads it at the start of every session (memory docs).

You don't have to write the first version yourself. The /init command analyzes the codebase and creates a CLAUDE.md with the commands and conventions it finds. If the file already exists, it suggests improvements instead of overwriting it. Claude Code also keeps its own notes from your corrections, called auto memory.

In our experience, the most useful notes are the ones the code can't show. Rules like "data fixes go through the portal, not the server" or "don't add this line to commit messages" stop the same mistake from happening twice.

What does a real task look like?

This site runs on Angular with server-side rendering, Spring Boot and PostgreSQL. The project started on September 11, 2026. Since then the repository has grown to nearly 150 commits, and we do our day-to-day development with Claude Code.

One example: Search Console showed 191 impressions from the US in a single day, with zero clicks. We asked Claude Code, in one sentence, to analyze it. It read the Search Console data in the browser, inspected the live pages' HTML and schema markup, and came back with a prioritized list.

The analysis showed that 95 percent of those impressions came from desktop and that part of them looked like automated queries. Claude Code then shipped the four items we picked (author pages, company profile links, an hreflang fix and IndexNow) in the same session. After each change it built the frontend, ran the backend tests and checked the result in a local preview. After the commit and push, GitHub Actions deployed the site in three to four minutes.

Why do permission modes matter?

Permission modes set what Claude Code may do without asking. In manual mode it asks before editing files and running commands. Plan mode is for exploring the code before changing anything. In auto mode, a separate safety check reviews actions in your place.

You switch modes with Shift+Tab in the terminal (permission modes). We also keep a few limits regardless of the mode. Claude Code never types passwords; we log in to Search Console and the portal ourselves. It has no direct access to the production server either: it prepares the commands that must run there, and we run them.

How do you catch mistakes?

Claude Code makes mistakes too; what matters is catching them before they ship. In the same session, an SEO check it had added counted Turkish sites on .com domains as foreign sources. It noticed this itself while reviewing a draft post, fixed it and confirmed the fix with a test.

It also worked the other way: while reading the code, it found an older bug where an English post's address changed with every translation. A changing address could have turned pages already indexed by Google into 404s. In both cases the same habit paid off: make it verify, and read the change.

How to use Claude Code: practical tips

1. Describe the task by the outcome you want, such as "find out why the US gets no clicks".
2. For larger work, ask for a plan first and see which files it will touch.
3. Make "build, run the tests, check the preview" part of the task.
4. Write recurring corrections into CLAUDE.md.
5. Watch the cost of automations that call the API. Our blog's automatic translation sent a paid request on every edit, so we made it on-demand.

Frequently asked questions

Is Claude Code free?

Usually not. Most surfaces require a Claude subscription or a pay-as-you-go Anthropic Console account. The official docs list the current options.

Where does Claude Code run?

In the terminal, in the VS Code and JetBrains extensions, in the desktop app and in the browser. They all use the same engine, so your CLAUDE.md files and settings work everywhere.

What should go into CLAUDE.md?

Build and test commands, coding conventions, the project layout and "always do X" rules. For long procedures that only matter for one task, the docs recommend separate mechanisms.

Can Claude Code deploy code on its own?

If you allow it, it can commit and push. We keep deployment behind an approval and give it no access to the production server.

The real answer to how to use Claude Code shows up in practice: a clear task, a good CLAUDE.md and verification at every step. For the credits used by Claude's cloud sessions, see our Claude session credit guide. If you want a similar development workflow for your own project, reach us through our web application development page.

Sources