I'd heard about Claude Code for months. A command-line tool that lets you code with AI. People raved about it. I kept scrolling.
The terminal was too intimidating.All that black screen and blinking cursor energy. I'm not a coder. I knew enough to know that command-line tools weren't built for people like me.
Then, one week ago, everything changed.
The First Wall: Copy-Pasting Code
I was deep in an n8n automation project. n8n is a workflow automation tool — think Zapier but open-source and way more flexible. I was building something complex, and Claude was helping me write the code.
Here's what that looked like:
- Ask Claude for code
- Copy Claude's response
- Paste into my n8n node
- Run it
- Get an error
- Copy the errorClipboard gymnastics. My least favorite sport.
- Paste back to Claude
- Get a fix
- Repeat
I was spending more time copying and pasting than actually solving problems.
So I asked ClaudeVery meta. I asked the AI how to collaborate with it better. And it told me.: "How can I work with you more efficiently?"
Claude told me about MCP servers — specifically the filesystem server. MCP stands for Model Context Protocol, and Anthropic announced it in late 2024. Think of MCP servers as bridges that let Claude interact with things outside the chat window.
The filesystem server gives Claude read/write access to a folder on your computer. You point it at your project folder, and now Claude can edit your files directlyYes, I triple-checked the permissions before trusting it with my project..
No more copy-pasting code. I describe what I want, Claude makes the changes, I see them instantly.
I still use Claude + MCP filesystem for my n8n projects. It works great for that.
But then I hit another wall.
The Second Wall: Copy-Pasting Terminal Commands
I started building this portfolio site. Different project, different friction.
Claude would tell me to run commands like pnpm dev, pnpm build, git add .I'm not exactly sure what all those commands mean. But they work., git commit, git push.
I'd copy each command, paste it into the terminal, run it, copy any errors back to Claude, get a fix, repeat. Sound familiar?
So I asked Claude again: "Is there a way you can just run these commands for me?"
That's when it told me about Claude Code.
The Toggle That Changed Everything
Here's what I didn't realize: Claude Code isn't just a separate terminal app anymore. In the Claude desktop app for Mac, you can switch from chat mode to Claude Code mode. Toggle it on, and the terminal interface becomes a chat window you recognizeSometimes the barrier isn't capability. It's just the interface..
Same power, way less intimidating than a raw terminal.
Claude Code can edit files and run terminal commands. When it needs to execute something, it asks for permission first. I'm careful about what access I allow — I read each prompt before approvingTrust, but verify. Especially when an AI wants to touch your terminal..
What One Week Looks Like
I'm not an expert on Claude Code. I'm sure I'll pick up more tricks. But here's what I can do after one week:
I built a portfolio site. Not by writing code — by describing what I wanted. Claude Code created the files, structured the project, styled the pages. When something looked off, I told it. It fixed it.
I wrote a blog post by talking. My first post, The Ceiling Is Gone, started as a rambling conversation. Claude refined it, picked out the blockquotes, formatted the metadata. The post sounds like me because we wrote it together.
I set up custom commands. Now when I type /blog, Claude Code knows my site structure, my voice, my formatting preferences. This post you're reading? That's how it started.
I added features I'd given up on. I wanted sidebar annotations — hover over text, see my inner thoughts in the margin. Years ago, I tried plugin after plugin on WordPress. They were buggy, slow, incompatible. I gave up. With Claude Code? A few prompts. Done.
The Part Nobody Talks About
I'm not going to sugar-coat this. YouTube tutorials love to show "one prompt, done!" That's not reality.
The hardest part was infrastructure setup:
- Localhost — running your site on your own computer (localhost:3000) so you can preview changes before they go live
- GitHub — think of it as Dropbox for code.Coders, forgive the oversimplification. Non-coders, you're welcome. Your local files sync to an online repository.
- Netlify — free hosting that watches your GitHub repo and automatically publishes changes
Setting this up took the longest. Understanding why things weren't working, troubleshooting errors I'd never seen before. But I wanted to learn as I built, so I asked Claude to explain things along the way. It patiently answered every question. #AlwaysLearning
Once the pipes are connected, everything flows. Claude Code talks to your local files, GitHub syncs them, Netlify publishes. You just... talk.
Where to Start
If this resonates, here's my advice: start a project BEFORE you're ready.
I didn't learn Claude Code from a tutorial. I learned because I was knee-deep in an n8n project, then a portfolio project, and the friction became unbearable. The obstacles taught me what I needed to learn.
Pick something you actually want to build — a blog, a portfolio, an automation, whatever. Start working on it even if you don't know exactly how to deploy it.
Get a Claude Pro account.No financial affiliation. I just think Claude has the best coding quality and most human-like writing. Although Claude, if you're listening — sponsorship welcome. Download the desktop app. Switch to Claude Code mode. When you hit a wall, ask Claude how to work with it better.
You might be surprised what it tells you.