I’ve been using Claude Code almost every day for the past five months. I built this website with it, prototyped a retirement financial planner, and wired together a voice-activated expense tracker. I don’t have a CS degree. Here’s how to get started yourself.

What Claude Code actually is

Claude Code lives in your terminal. Unlike the Claude website or app, it can see your files, create new ones, edit existing ones, and run commands on your machine. It’s not a chatbot. It’s an assistant that can actually do work in your project folder.

You don’t need to know how to code to use it. You do need to be comfortable typing commands and reading what comes back.

What you need

  • A computer (Mac, Windows, or Linux)
  • An internet connection
  • A project folder (any folder with files you’re working on)
  • An Anthropic account at claude.ai

You don’t need GitHub, programming languages, or previous AI experience.

Three ways to pay

API Key (pay per use): Load credit and pay for what you use. Like a prepaid phone. Start with $10-20 to test. Best for trying it out.

Claude Pro ($20/month): A subscription with a set amount of usage per month. Best for light, predictable use.

Claude Max ($100 or $200/month): High volume. Best for daily heavy use. I use the $100 tier.

Installation in 4 steps

  1. Install Node.js from nodejs.org (download the LTS version, run the installer)
  2. Open your terminal (Mac: Cmd+Space, type “Terminal”. Windows: search “PowerShell”)
  3. Run: npm install -g @anthropic-ai/claude-code
  4. Navigate to your project folder with cd, then type claude

That’s it. Claude is now running and can see every file in your project folder.

Your first interaction

Try asking:

What files are in this project? Give me a summary of what you see.

Claude will scan your folder and describe what it finds. This isn’t a generic AI response. It’s specific to your actual files.

Try creating something:

Create a file called test-note.md with a summary of this project based on what you can see.

Claude will ask permission before creating or editing any file. You approve each action with y. Nothing happens without your explicit yes.

The 8-layer system

This is the part most people skip, and it’s the part that makes Claude Code actually useful long-term. Think of it as progressive complexity. Start simple, add layers when you feel friction.

LayerWhat it isComplexity
FoundationClaude Code installedBeginner
1. CLAUDE.mdA rules file Claude reads every sessionBeginner
2. CLAUDE.local.mdPrivate notes Claude can referenceBeginner
3. Slash commandsReusable workflows you trigger with /Intermediate
4. HooksAutomated triggers on specific eventsIntermediate
5. Memory systemPersistent context across sessionsIntermediate
6. MCP + searchSemantic search over your filesIntermediate
7. SkillsAdvanced reusable behaviorsIntermediate
8. Permissions + settingsFine-tuning autonomy and accessAdvanced

Start with the Foundation and Layer 1. Add the rest when you feel friction, not on a schedule.

Layer 1: The CLAUDE.md file

This is the single most impactful thing you can do after installation. It’s a markdown file at the root of your project folder. Claude reads it automatically at the start of every session. It contains your rules.

Ask Claude to create one for you:

Look at my project folder and create a CLAUDE.md with:
1. A folder map showing where different files should go
2. File naming conventions based on what you see
3. Any patterns you notice
4. 5-10 rules for how you should work in this project

Start with 10-15 lines. A mature CLAUDE.md might grow to 400+ lines over months. Add rules when you notice Claude doing something you don’t want.

What I’ve built with it

  • This website. I described what I wanted. Claude handled the Astro framework, CSS, and deployment config. I made the design and content decisions.
  • A retirement financial planner prototype. Input your numbers, get projections. Claude wrote the calculation logic while I defined the requirements.
  • A voice-activated expense tracker. I wired together Telegram, ChatGPT, Make.com, and Google Sheets so I could log expenses by talking into my phone. Configured the entire automation manually.
  • A career opportunity scanner and resume system. It monitors company career pages, scores leads against my profile, tailors resumes to each job description, and generates application materials in batches.

None of these required me to write code from scratch. I described what I wanted, reviewed what Claude produced, and iterated until it worked.

Honest limitations

  • Claude sometimes gets things wrong. You need to review its output, especially for anything involving numbers or logic.
  • Large projects can hit context limits, though the 1M context window on Opus 4.6 handles most sessions well. When you do hit the limit, Claude auto-compacts the conversation and keeps going.
  • It’s not free. Heavy daily use on the API can cost $20-50/month. The Max subscription ($100 or $200/month) gives you predictable costs. I use the $100 tier.
  • You still need to know what you want. Claude is great at implementation, but you have to bring the vision and the judgment.

Where to go deeper

I put together a full starter kit with slides, a cheat sheet, and a detailed walkthrough. If you want a copy, reach out.