Claude Code tutorial: your first AI agent in 30 minutes

Claude Code vs your current options

What Claude Code actually is

Claude Code is an agent, not a chatbot. That difference matters more than anything else in this guide.

A chatbot (ChatGPT, Claude.ai, Gemini) waits for you to type something. You type, it responds. You type again, it responds again. The work always flows through you, and you can't step away mid-task.

An agent works differently. You give it a task, it reads files on your computer, writes new ones, sends requests to external services, and hands you a finished result. You don't babysit it while it works. You review the output afterward.

Claude Code uses Anthropic's Claude model as its brain but runs as an agent in a terminal window on your computer. When you trigger a task, Claude Code reads the instruction you wrote, follows each step in order, and gives you something to review. You're not doing the work. You're reviewing it.

The blog post you're reading right now was written, formatted, and published to this site by a Claude Code task. No human typed it into a content management system. A plain-English instruction file described what to do, Claude Code ran it, and the result landed on your screen.

If you want the full conceptual explanation before diving into setup, the What Is Claude Code? guide covers the agent model in depth. This tutorial skips straight to getting it working.

What you need before you start

Three things, all free or nearly free to get:

That's the full list. No developer experience. No existing project. No other software beyond a web browser for the API key setup.

Step 1: install Claude Code

Open your terminal. On Mac, search for "Terminal" in Spotlight. On Windows, open Command Prompt or PowerShell. On Linux, you already know where it is.

Paste this single line and press Enter:

npm install -g @anthropic-ai/claude-code

This installs Claude Code globally on your computer. When it finishes, set your Anthropic API key so Claude Code knows which account to bill:

claude config set apiKey sk-ant-your-key-here

Replace the placeholder with the key you copied from console.anthropic.com. You only do this once. Verify the install worked:

claude --version

If a version number prints, you're ready. If anything goes wrong, paste the error message into a new Claude Code session and it will diagnose the issue and tell you exactly what to fix. That's one of the useful things about using an AI agent for setup problems.

Step 2: your first conversation with Claude Code

Navigate to any folder on your computer. It can be empty or an existing project folder. In the terminal, run:

claude

Claude Code opens an interactive session with a prompt where you can type. Start simple to confirm it's working:

What files are in this folder?

Claude Code lists them. Nothing impressive yet — just confirming the agent can read your file system. Now give it something to create:

Create a file called contacts-to-follow-up.txt and list five questions I should ask a new sales prospect in our first call.

Open the file. You'll find five well-structured questions, written into a real file on your computer. That's the basic loop: describe an outcome, Claude Code creates it in your files. Close the session with /exit when you're done.

The interactive session is useful for one-off jobs. For recurring work, you want a skill file.

Step 3: write a skill file and run it repeatedly

A skill file is a plain-English document (a .md text file) that describes a multi-step task. You write it once. Claude Code runs it every time you need.

Here's what a simple skill file looks like. Save this as weekly-crm-update.md in any folder:

# Weekly CRM Summary

1. Read the file called contacts.csv in this folder.
2. Count how many contacts have "New" in the Status column.
3. List their names and email addresses.
4. Write a plain summary to a file called weekly-update.txt.
5. Tell me when it is done.

Start a Claude Code session in that folder and run the skill by typing /weekly-crm-update (the filename without the .md extension). Claude Code reads the instruction, carries out each step in order, and creates the output file. You review the result. If anything looks off, update the instruction and run it again.

This is the whole pattern. Write the instruction once; the agent runs it on demand. But the real payoff isn't running it yourself every Monday — it's never having to remember at all.

Step 4: put your skill on a schedule

So far you've been triggering the skill by hand. The last step is to hand off the timing too — tell Claude Code when to run it, in the same plain English you used to write the skill. Start a session and type something like:

Schedule my weekly-crm-update skill to run every Monday at 9am.
Run it on its own and leave the summary in this folder for me to review.

Claude Code sets up a recurring schedule — it calls these "routines" — that runs in the background, in the cloud, so it fires even when your laptop is closed. You don't write a single line of cron syntax or touch a settings file; you describe the timing and Claude Code handles the wiring.

From then on, the summary is simply waiting for you every Monday morning. You still keep the final say — the agent does the work and leaves the result for you to check, not publish blindly. And because it's plain English, changing it is just as easy: "make that every weekday instead," or "pause the weekly-crm-update schedule," and it's done.

That's the full arc of an AI agent: write the instruction once (Step 3), then decide whether you trigger it or it triggers itself (Step 4). Everything after this is just adding more skills.

Step 5: skip the writing with a pre-built skill library

Writing your own skill files from scratch takes practice. The faster path: start with someone else's, already tuned to your exact use case.

SoloStack ships a library of over 80 pre-built skill files covering the most common small-business operations. Lead enrichment and scoring. Email sequence automation. SEO blog publishing. LinkedIn carousel generation. Weekly CRM reports. Webinar rotation. Each skill is a plain-English file you can read, understand, and adapt without technical knowledge.

What SoloStack runs on Claude Code: real examples

To make this concrete, here is what Claude Code handles in the SoloStack business every week. No human does any of this manually.

The pattern across every function: Claude Code handles execution, filing, drafting, and sending. The operator handles the judgment calls — is this post actually good? is this lead a real fit? That split is what makes it sustainable. The AI does the volume work; you do the 10% that actually requires your expertise.

Pick one task and build from there

The mistake most people make is trying to automate everything at once. Pick one recurring task that costs you real time. Run Claude Code on that task until the output is reliable. Then add the next one.

Good first automations for small businesses:

• A weekly summary of your inbox: who responded, who didn't, what needs a follow-up
• A first draft of a blog post from a headline and three bullet points you supply
• A standardised client onboarding email filled in from a new contact's details in your CRM
• A list of your top leads by last-contacted date, written to a file every Monday morning

None of these require writing any code. They require writing a clear instruction in plain English, which is the same skill you use to explain a task to a new team member. The AI just happens to follow instructions without forgetting them.

SoloStack can help you get Claude Code running for your specific business — from the first install to your first automated workflow, built live with you in the workshop. We cover the setup, the skill library, and the first real task that saves you time before the session ends.