Carlos Imlau · Blog

Claude Code without knowing how to code: how I started and what I learned

Carlos Imlau, AI Product Builder · Certified trainer (CCP) · AI applied to work ·

Summary

You can use Claude Code without knowing how to code: what you need to learn is directing the AI, not writing code. With a small problem, a context document and step-by-step requests, I rebuilt Tapinfy in about a month.

  • Claude Code works inside your project folder: it reads, writes and runs commands with your permission
  • You need to know what you want, test everything and lose the fear of the terminal
  • Ask in stages and test each one before moving on
  • Save versions with Git and publish early

Can you use Claude Code without knowing how to code? Yes. What you need to learn is not writing code, it is directing the AI: saying what you want, checking the result and deciding what stays. The AI writes the code; you decide the product. I am not a programmer, and that is how I rebuilt Tapinfy, my booking app, in about a month.

I started like many people: copying code from ChatGPT, pasting it into the editor and watching it break. Today I work with seven terminals open, one project each, with Claude working in every one. This article explains what changed.

What is Claude Code?

Claude Code is Anthropic's coding agent. Instead of answering in a chat, it works inside your project folder: it reads the files, writes and edits code, runs commands and asks for your permission before touching anything important. You can use it in the terminal or in the desktop app.

The difference from a regular chat is big. In a chat, you copy and paste. With an agent, the AI sees the whole project and works on it, and your job becomes asking, checking and approving.

Do I need to know how to code to use Claude Code?

No. But you need three things people rarely mention:

Como eu uso IA para criar os vídeos de divulgação do meu app
Como eu uso IA para criar os vídeos de divulgação do meu app

How to start with Claude Code in 5 steps

  1. Pick a small, real problem. Something of your own that you would use tomorrow. Your first project is not your life's product.
  2. Create the project folder and a context document. A simple file with what it is, who it is for, what is out of scope and the rules. The AI reads it at every step.
  3. Ask in stages. First the structure, then one screen, then login. Asking for everything in one message is the fastest route to a project nobody understands.
  4. Test each stage before moving on. If something broke, send the exact error. If it looks bad, say what you want instead of "make it better".
  5. Save versions and publish. Use Git so you can go back, and publish early, even if only for yourself.

What mistakes did I make at the start?

Claude Code or ChatGPT: which one should I use?

Both have a place. I use chat to think, write and plan, and the agent to work on the files. In my current workflow, Claude is the AI I use most and Codex comes in as a code reviewer. Pick one agent, learn to work with it well, and only then compare.

Frequently asked questions

Is Claude Code paid?

Yes, it needs a paid Anthropic plan or API credits. Prices change often, so check the current figures on the official website before you start.

How long until I can publish something?

It depends on what you want to build. A simple page can go live the same day. Tapinfy, which has accounts, a calendar and clients, I rebuilt from scratch in about a month.

Do I need a powerful computer?

No. The heavy work runs on the AI's servers. A normal computer with internet is enough.

Where can I learn the full step by step?

In the SaaS Starter (in Portuguese) I put the template and guide I use to build with Claude Code or Codex, from the first screen to launch. If you prefer guidance, I offer one-to-one mentoring to build your product with me.