Claude Code: first steps without being a developer

Until now you have used AI inside a chat window: you paste, it answers, you copy. This module changes where the work happens. Claude Code is an agent that lives in a folder on your computer, reads the files in it, creates or edits whatever is needed and checks that the result works. You do not need to know how to program to use it. You need to describe what you want and read what it gives you, which is exactly what you practiced in Parts 1 and 2.

We go in this order: what a coding agent is and what it can do for a business, where Claude Code runs and which plan you need, the terminal in ten minutes, installation and the first session, and your first real tool. If you are a developer, skip the terminal section; the rest applies just the same.

What is a coding agent, and how is it different from a chat?

In a chat, the model can only answer with text. A coding agent has tools: it can read a file, search inside a folder, write or edit files, run a command and read the output. That turns it from an adviser into someone who does the work. The Claude Code documentation describes how it works as a loop of three phases that blend together: gather context, take action and verify results.

1. 1Gather context. It reads the files in the folder you pointed it at, searches for what is relevant and works out how the project is put together. You do not paste anything: it reads files as it needs them.
2. 2Take action. It creates or edits files and runs commands (start a test server, run a script, install a library).
3. 3Verify. It checks that what it did works: runs tests, runs the thing again, reads the errors and fixes them. Then it repeats the loop until it decides the task is done, or until you interrupt it.

It does all of this inside a folder you choose. That folder is its world: it sees what is inside, it can run the commands you could run, and if the folder is a repository, it sees the change history. Outside the folder it touches nothing without asking. That boundary is your first safety tool.

What it is good for if you run a business and do not program:

• Small internal tools. A page that calculates a quote with your rules, a form that builds a contract from a template, a dashboard that reads an export from your system and shows what you care about.
• File chores. A script that cleans your spreadsheet export every Monday, renames two hundred photos by what is in them, or turns twenty PDFs into a table.
• A simple website. A landing page for a new service, with your copy and your photos, that you can publish without paying an agency for every change.
• Understanding what you already have. If someone built you a system, Claude Code explains in plain language what each part does. That is the question you will start with.

The automations from module 8 live here too: what you did there with n8n or scheduled tasks, with Claude Code you can build as a small program that runs on its own. We come back to that in module 15.

Where does Claude Code run, and which plan do I need?

As of October 2026, the official documentation lists four main surfaces: the terminal (the full version, and the one in this module), extensions for code editors (VS Code, Cursor and the JetBrains IDEs), a desktop app for Mac and Windows that includes Claude Code with nothing else to install, and the web at claude.ai/code, where the session runs in the cloud. They all use the same engine: what you learn in one carries over to the others.

On plans: according to the documentation, Claude Code requires a Claude Pro, Max, Team or Enterprise plan, or a Console account (the API one, with prepaid credits). The free plan does not include it. If you already pay for Pro to use the chat, you already have it; it is not a separate product.

What do I need to know about the terminal? (ten minutes)

The terminal is a window where you type an instruction, press Enter and see the result as text. It has no buttons and you cannot click on things: you move with the keyboard. It looks like something from another era, but it is the natural home of programming tools, and Claude Code is a CLI, a program you drive by typing. You only need five ideas.

1. 1Open it. On a Mac: Cmd + Space, type "Terminal", Enter. On Windows: Win + X and choose Windows PowerShell or Terminal. You will see a line with a blinking cursor: that is where you type.
2. 2Know where you are. The terminal is always "standing" in a folder. Type `pwd` and Enter to see which one (it works in Windows PowerShell too). That path, something like `/Users/ana/Documents` or `C:\Users\ana\Documents`, is the folder’s full address.
3. 3See what is there. Type `ls` to list the files and folders where you are (in PowerShell `ls` works too; `dir` is the classic Windows equivalent).
4. 4Move around. `cd Documents` goes into the Documents folder; `cd ..` goes up one level; `mkdir quoter` creates a new folder called quoter. If a name has spaces, put it in quotes.
5. 5Run and stop. A command runs when you press Enter. If something keeps running and you want to stop it, Ctrl + C. To close the terminal, type `exit`.

How do I install Claude Code and run my first session?

These are the official steps as of October 2026 (sources at the end). If something fails, the installation guide in the documentation has a list of errors with their fixes, and if you have never opened a terminal before, Anthropic has a beginner’s guide linked below.

1. 1Install. Open the terminal and paste a single line. On Mac or Linux: `curl -fsSL https://claude.ai/install.sh | bash`. In Windows PowerShell: `irm https://claude.ai/install.ps1 | iex`. It shows no progress while it downloads: wait until you see the success message. On Windows, the documentation also recommends installing Git for Windows (a normal installer) so Claude Code has all of its tools.
2. 2Check. Open a new terminal and type `claude --version`. If it prints a version number, you are set. If it says the command is not recognized, the installation guide explains how to fix the path.
3. 3Go into a folder and start. `cd` into the project folder (or create an empty one with `mkdir`) and type `claude`. The first time it asks you to sign in: the browser opens, you log in with your Claude account and come back to the terminal. You will not have to repeat it.
4. 4Talk to it. The Claude Code prompt appears with the version, the model and the current folder. You type in plain language and press Enter. `/help` shows the commands; Esc interrupts whatever it is doing; `exit` (or Ctrl + D twice) closes the session.

The first session has a fixed script, and it is worth following even if you are in a hurry. First, have it explain the folder: if it is a project someone built for you, this is the X-ray you were never given; if it is an empty folder, it will tell you it is empty, and that is information too. Second, ask for a small, visible change, the size of "create a page that says hello". Third, read what it changed before accepting it.

Explain this project to me as if I could not program. I want to understand:

1. What it does, in one sentence I could say to a customer.
2. What is in each main folder, one line each, no jargon.
3. Where the copy, prices or data that I would change most often live.
4. Which commands are used to test it and to publish it, if you can work that out from the files.
5. Three things that look fragile or that I should know before asking for changes.

Do not change any file yet. If something cannot be known from reading the files, tell me instead of guessing.

About permissions: Claude Code has modes. In manual mode it asks before editing a file or running a command and shows you the exact change (the lines it removes and the lines it adds) so you can approve or reject. As of October 2026, recent versions start in an auto mode where a second model reviews actions in the background and blocks the risky ones instead of asking you; Shift + Tab switches modes at any time. For your first sessions, switch to manual: approving every change is slow, but it is how you learn to read what it does. Also, before editing a file, Claude Code saves a snapshot: if something went wrong, Esc twice rewinds to an earlier point.

How do I build my first tool without programming?

The difference between a toy and a useful tool is not in the code; it is in the description. A request like "make me a quote calculator" produces something generic you will correct ten times. An eight-line spec produces something that looks like what you need from the first version. It is the Description skill from module 4 applied to building, and the whole subject of module 17.

The mini spec we recommend for a one-page tool has four parts: who uses it (and when), what goes in (the data the person types or pastes), what comes out (what the tool shows or generates) and the rules (how it is calculated, which special cases exist, what it must not do). Ask for the simplest version that works, in a single HTML file that opens in the browser with a double click, no servers and no databases. Then iterate.

Build the first version of [TOOL: for example, "an equipment rental quote calculator", "a welcome message generator", "a commission calculator"].

Who uses it: [ROLE AND MOMENT: for example, "the salesperson, while talking to the customer on WhatsApp"].
What goes in: [LIST OF FIELDS: for example, "number of days, equipment type (three options), whether delivery is needed"].
What comes out: [WHAT IT SHOWS: for example, "the itemized total and a text ready to copy and paste to the customer"].
Rules: [HOW IT IS CALCULATED AND THE SPECIAL CASES: for example, "daily rate by type; 10% off from 7 days; delivery costs 40,000 inside the city; never show prices without VAT"].

Constraints: a single HTML file that opens in the browser with a double click, no server and no database, in Spanish, working on a phone. Start with the simplest version that meets the rules. Before writing code, tell me in five lines how you will do it and ask me whatever is unclear.

When it finishes, it will tell you how to open the file. Open it, use it the way the real person would, with normal cases and odd ones (zero days, a negative number, an empty field), and come back with what you saw: "with zero days it shows a negative price", "the customer text should be friendlier", "add a weekend option". Each round is a short message; do not start over. And whenever it makes a change you do not understand, use the third prompt.

Explain what you changed in this last round and how I test it:

1. List the files you touched and, for each one, what changed and why, in plain language.
2. Tell me exactly what I should do to test it myself (what to open, what to type, what I should see).
3. Tell me which case you did NOT test, or which assumption you made that I should confirm.
4. If anything you did could break easily with a future change, warn me.

If you want the official walkthrough, Anthropic Academy’s free Claude Code 101 course (about twelve short lessons) covers the same ground in more depth; it is aimed at developers, but after this module you can follow it.

In your first session you will have noticed something: every time you start, Claude Code begins from zero, not knowing what your project is or which rules you have. The next module, "CLAUDE.md, memory and context management", fixes that: you teach it your project once, in a file, and you learn to keep sessions clean so it does not get lost in long tasks.