Claude Code for Beginners: Your First Project in 30 Minutes

Your Claude Code first project does not need to be impressive. It needs to be instructive. Right now you have a working CLI tool and no clear picture of how it behaves when you give it a real task.

This walkthrough takes you from an empty folder to a working result in 30 minutes. You will leave with a clear understanding of how to prompt, review, and iterate on Claude Code output.

Key Takeaways

• CLAUDE.md is not optional for beginners: Writing even a minimal CLAUDE.md before your first task gives Claude Code the project context needed to produce relevant output.
• Specific prompts produce usable output: "Build me a to-do app" produces mediocre results. A prompt naming specific files, behaviours, and constraints produces something you can evaluate.
• Iteration is the core skill: Expect 3 to 5 follow-up instructions after the first response. This is normal, not a failure.
• Plan mode is the safest start: Before writing any files, Claude Code shows you what it plans to do. Reviewing the plan prevents unwanted changes.
• Git is your safety net: Run git init before starting. This lets you revert any file Claude Code creates or modifies.

AI App Development

Your Business. Powered by AI

We build AI-driven apps that don’t just solve problems—they transform how people experience your product.

Let's talk

Before You Start: What Do You Need?

You need Claude Code installed and authenticated, a terminal, and a text editor. No prior AI coding tool experience is required for this tutorial.

If setup is not complete yet, the prerequisites are straightforward to clear before starting the tutorial steps.

• Installation and authentication: If you have not done this yet, follow the guide to install and authenticate Claude Code, or the authentication setup guide if installation is complete but auth is not.
• A terminal: Terminal on macOS, Ubuntu in WSL on Windows, or any Linux terminal all work without extra configuration.
• VS Code as your text editor: It shows file changes in real time alongside the Claude Code terminal session, which is useful for a first run.
• No prior experience needed: This tutorial assumes this is your first AI coding tool session. Every step is shown explicitly.

Once you have those four things confirmed, you are ready to start.

Step 1: How Do You Create Your Project Folder?

Create a new directory, navigate into it, and initialise git immediately. That three-step sequence gives Claude Code a defined workspace and gives you a safety net from the start.

The project folder is Claude Code's working directory. It reads, writes, and executes within this folder by default.

• Create and enter the directory: Run mkdir my-first-claude-project && cd my-first-claude-project to set up and enter your workspace in one command.
• Initialise git immediately: Run git init before anything else. This creates the safety net that lets you revert any change Claude Code makes that you did not intend.
• Create a starting file: Run touch index.html for a web project. Claude Code performs better when the project is not completely empty.
• Stick with a web project for this tutorial: HTML, CSS, and JavaScript produce visible output quickly and require no build tools or dependency installs.

With the folder created and git initialised, you are ready to start a Claude Code session.

Step 2: How Do You Start a Claude Code Session?

Run claude in your project directory. Claude Code starts in interactive mode and waits for your first message. It reads your project files automatically before you type anything.

The interface is a conversational CLI, not a traditional command-line tool. It shows your current directory, your conversation history, and a text input line.

• Launch from the right folder: Run claude from inside your project directory. If you launch from your home directory, Claude Code will not find your project files.
• Type /help first: This shows all available slash commands and is the fastest way to orient yourself to what Claude Code can do.
• Enable plan mode with Shift+Tab: For a first session, plan mode ensures Claude Code shows you its intended actions before making any file changes. Enable it before your first prompt.
• Claude Code scans your directory on start: It builds context from your existing files before you type. This is why starting in the correct folder matters.

With plan mode enabled, you are ready to write your CLAUDE.md before giving Claude Code its first task.

Step 3: How Do You Write Your First CLAUDE.md?

CLAUDE.md is a plain text file Claude Code reads automatically at the start of every session. It provides persistent project context so you do not have to re-explain your project each time you launch.

Create it in your project root: touch CLAUDE.md, then open it in VS Code. A minimal file for a beginner project needs only three things.

• What the project is: One sentence describing the purpose. Claude Code uses this to make sensible default decisions throughout the session.
• The tech stack and constraints: For example, "plain HTML, CSS, JavaScript only, no frameworks, no build tools." This prevents Claude Code from reaching for libraries you do not want.
• Naming conventions: Something like "kebab-case for file names." Claude Code gets naming conventions wrong without explicit guidance.

A minimal working CLAUDE.md for this tutorial looks like this:

# My First Project A simple to-do list web app. Stack: plain HTML, CSS, JavaScript — no frameworks, no build tools. File naming: kebab-case.

Without a CLAUDE.md, Claude Code produces correct but generic output. With it, responses are scoped to your actual project. For advanced patterns across multiple projects and team settings, the full CLAUDE.md reference covers all options.

Step 4: How Do You Give Claude Code Its First Task?

Write a prompt that names a specific file, a specific behaviour, and specific constraints. Vague prompts produce output you cannot evaluate. Specific prompts produce output you can iterate on.

The difference between a useful first prompt and a frustrating one is specificity. Here is that difference shown directly.

• Vague prompt: "Build me a to-do app." Claude Code will produce something, but it will make assumptions about frameworks, structure, and styling that may not match what you want.
• Specific prompt: "Create index.html with a to-do list. Add items via a text input and button. Store the list in localStorage so it persists on refresh. Style with style.css, clean and minimal, no frameworks."
• Review the plan before approving: In plan mode, Claude Code displays what it intends to do before acting. If something looks wrong, type a correction before approving.
• Open the files before asking for changes: After Claude Code finishes, open the files in your text editor and in a browser. Evaluate what it produced before requesting any modifications.

The three-part prompt structure is: what to create, what it should do, and what constraints apply. Use that structure on every first prompt.

Step 5: How Do You Review Output and Iterate?

Open the generated files and evaluate them against what you asked for, not against a finished product. Then give specific corrections, not restarts. That is the entire iteration loop.

Effective iteration is a learnable skill. The difference between a correction that works and one that does not comes down to specificity.

• Specific corrections outperform restarts: "The button is not aligned with the input. Fix the layout in style.css so they sit on the same line" is far better than "the styling is wrong, try again."
• Use /undo if a change made things worse: This reverts the most recent file change Claude Code made, without affecting earlier changes in the session.
• Ask Claude Code to explain its reasoning: "Why did you structure the JavaScript this way?" produces a walkthrough of the decision, which is useful for both learning and targeted fixing.
• Expect 3 to 5 rounds before a satisfying result: Each round you are getting more specific about what you want. This is the skill that makes Claude Code genuinely useful at larger scale.
• Commit after iterating: Run git add . && git commit -m "first claude code session" to checkpoint your progress before the next session.
• Reference the full command set: For every command available during iteration, the complete CLI commands reference covers flags, shortcuts, and slash commands.

The iteration loop from this tutorial is the same loop used on production codebases. The project gets more complex; the approach stays the same.

What Are the Most Common Beginner Mistakes to Avoid?

The mistakes that derail first sessions are predictable and avoidable. Almost all of them come from skipping setup steps or using prompts that are too broad to evaluate.

Knowing these in advance prevents the specific frustrations that cause people to give up on Claude Code after one session.

• Starting without git: Claude Code can create and overwrite files in auto mode without warning. git init before every session is your only safety net.
• Prompts that are too broad: "Build me a web app" produces code you cannot evaluate or iterate on meaningfully. Every first prompt should name specific files, behaviours, and constraints.
• Accepting the first output without reviewing it: Claude Code produces plausible-looking code that may have logic errors or missing edge cases. Always open and read the files before marking a task complete.
• Restarting instead of correcting: If the first output is wrong, correct the specific element that is wrong. Restarts lose the accumulated context from the conversation.
• Launching Claude Code in the wrong directory: If you run claude from your home directory, it will not find your project files or CLAUDE.md. Always cd into the project folder first.
• Skipping plan mode on a first session: Auto mode gives beginners no visibility into what Claude Code is about to do. Enable plan mode with Shift+Tab until you understand how it behaves on your project type.

Running into these once is understandable. Running into them repeatedly means a setup step was skipped.

What Should You Build Next?

The tutorial project was a single-file web page. Claude Code handles significantly more complex work once you understand the iteration pattern. Your second project is where the approach starts to feel natural.

The prompt specificity and iteration habits from this tutorial apply directly to more complex projects.

• Add a backend to your to-do app: A Node.js Express API connected to your existing frontend shows what multi-file iteration looks like, without requiring a new project.
• Convert an existing static page: Take an HTML page you already have and ask Claude Code to add interactivity, extract components, or write tests for its JavaScript.
• Full-stack at larger scale: A step-by-step walkthrough of how to build a full-stack app with Claude Code covers multiple files, a backend, and a database using the same iteration approach.
• A more experimental working style: For a higher-level, flow-based way of working where you describe intent and let Claude Code figure out structure, the vibe coding approach guide covers that pattern.

The skills compound. A more complex project does not require a new approach, only more specific prompts and more deliberate iteration.

Conclusion

Claude Code for beginners is not about getting the first prompt right. It is about building the iteration loop: write a specific prompt, review what comes back, correct the specific element that is wrong, and repeat.

That loop, applied on a simple project today, is the same loop that makes Claude Code reliable on a production codebase. Take what you built here and give it one useful extension before your next session. The second task is where the pattern clicks.

AI App Development

Your Business. Powered by AI

We build AI-driven apps that don’t just solve problems—they transform how people experience your product.

Let's talk

Ready to Use Claude Code on a Real Project, Not Just a Tutorial?

Getting comfortable with the iteration loop on a tutorial project is one thing. Knowing how to integrate Claude Code into an existing codebase or team workflow is a different challenge entirely.

At LowCode Agency, we are a strategic product team, not a dev shop. We work with teams adopting Claude Code for production development, helping them go from first sessions to a consistent, reliable workflow across their real projects.

• Workflow integration: We map your existing development process and identify where Claude Code fits without disrupting your current review and deployment pipeline.
• CLAUDE.md configuration: We write and test CLAUDE.md files for your specific stack so Claude Code produces consistent, on-convention output from the first session.
• Prompt pattern design: We develop the prompt templates your team uses repeatedly so every developer gets predictable results without needing to learn by trial and error.
• Codebase onboarding: We configure Claude Code to understand your existing architecture, naming conventions, and architectural rules before anyone on your team starts using it.
• Plan mode and safety protocols: We establish the approval workflows and git practices that prevent Claude Code from making unwanted changes in your production codebase.
• Iteration training: We run working sessions with your developers so they build the specific correction habits that make Claude Code useful rather than unpredictable.
• Full product team: Strategy, architecture, development, and QA from a team that has integrated Claude Code into real product workflows, not just tutorial projects.

We have built 350+ products for clients including Coca-Cola, American Express, and Medtronic.

If you want Claude Code working reliably on your real projects, talk to our team.