Getting Started with Claude Code

1The AI Development Revolution

What is Claude Code?

Claude Code is a command-line tool made by Anthropic (the company behind Claude) that lets you build software by having a conversation. Instead of writing code yourself, you describe what you want, and Claude writes it for you.

Think of it like having a very skilled programmer sitting next to you who:

• Never gets tired or frustrated
• Knows almost every programming language and framework
• Can explain what it's doing in plain English
• Works at the speed you can describe your ideas

How AI Changed Software Development

The old way (pre-2023):

1. Have an idea
2. Spend months/years learning to code
3. Build the thing (slowly, with lots of frustration)
4. Debug for weeks
5. Maybe ship something

The new way:

1. Have an idea
2. Describe it to Claude Code
3. Review what it built
4. Ask for changes
5. Ship it

2Why Python?

Python Reads Like English

Compare these two snippets that do the same thing:

Python:

if user_age >= 18:
    print("You can vote")
else:
    print("Too young to vote")

Python strips away the ceremony. You can often read Python code and understand what it does, even if you've never programmed before.

The Ecosystem

Python has libraries (pre-written code packages) for almost everything:

• anthropic - Talk to Claude
• openai - Talk to GPT
• streamlit - Build web interfaces
• supabase - Connect to databases
• pandas - Work with data (like Excel on steroids)

The Excel Connection

You already think in a way that maps to Python:

3The Development Mindset

The Four-Step Cycle

Step 1: Research

Before building anything, understand what you're building.

• What problem are you solving?
• Has someone solved this before?
• What tools/libraries exist for this?
• What are the common pitfalls?

Step 2: Plan

Break the big thing into small things. Don't try to build everything at once.

Example plan for a chatbot:

1. Create a basic Streamlit page with a text input
2. Connect to Claude API and get a response
3. Display the response on the page
4. Add conversation history
5. Save conversations to Supabase
6. Deploy to Railway

Step 3: Execute

Build one piece at a time. Take the first item from your plan and build it. Nothing else.

When working with Claude Code, be specific:

• Bad: "Build me a chatbot"
• Good: "Create a Streamlit page with a text input field and a submit button. When the user clicks submit, print whatever they typed."

Step 4: Test

Verify it works before moving on.

1. Run the code
2. Try to break it (enter weird inputs, click buttons twice, etc.)
3. Fix any issues
4. Then—and only then—move to the next piece

4Your Toolkit

GitHub: Your Code's Home

GitHub is where your code lives. Think of it as Google Drive for code, but with superpowers.

What is version control? Remember how Google Docs shows you revision history? Git (the technology) and GitHub (the website) do this for code—but better. Every save is called a "commit," and you can see exactly what changed in each one.

Supabase: Your Database

A database is a super-powered spreadsheet in the cloud.

Supabase gives you PostgreSQL (the most respected database), plus authentication, storage, and real-time capabilities.

Railway: Your Server

Your laptop can run code, but it's not always on. A server is a computer in a data center that runs 24/7, has a public address anyone can access, and handles multiple users simultaneously.

Railway watches your GitHub repo. When you push new code, it automatically redeploys. Magic.

Streamlit: Your User Interface

Streamlit lets you build web apps entirely in Python:

import streamlit as st

st.title("My First App")
name = st.text_input("What's your name?")
if st.button("Say Hello"):
    st.write(f"Hello, {name}!")

That's it. That's a complete web app. No HTML, no CSS, no JavaScript.

Logfire: Your Monitoring

When your app runs on a server, you can't see what's happening. Logfire lets you see logs, track errors, measure performance, and get alerts when something breaks.

5Setting Up Your Environment

Step 1: Install a Code Editor

Download VS Code from code.visualstudio.com and install the Python extension.

Alternatively, try Cursor from cursor.com — it's VS Code with built-in AI features.

Step 2: Install Claude Code

# Install Node.js first if you don't have it
brew install node  # Mac
# For Windows, download from nodejs.org

# Install Claude Code
npm install -g @anthropic-ai/claude-code

# Verify installation
claude --version

Navigate to any folder and run claude. It will ask for your Anthropic API key from console.anthropic.com.

Step 3: Install Python (with Claude Code's help)

Ask Claude Code:

"Help me install Python 3.11 on my Mac using pyenv. Walk me through each step and verify it's working."

Step 4: Create Platform Accounts

• GitHub: github.com
• Supabase: supabase.com
• Railway: railway.app
• Logfire: pydantic.dev/logfire
• Anthropic: console.anthropic.com

Step 5: Project Setup (Let Claude Code do it)

Ask Claude Code:

"Set up the basic project structure for a Python AI project. I need requirements.txt, .env.example, .gitignore, and an empty app.py file."

6Understanding AI & LLMs

What is an LLM?

LLM stands for Large Language Model. When you type "The capital of France is", the model predicts that "Paris" is the most likely next word. It does this based on patterns it learned during training.

That's all it does: predict the next token. But when you do this really well, at massive scale, something remarkable happens—the model can answer questions, write code, and have conversations.

The Major Players

Note: AI models update frequently! Check provider docs for the latest.

Claude (Anthropic)

Best for: Coding, long documents, reasoning, following complex instructions

ChatGPT (OpenAI)

Best for: General tasks, embeddings, broad knowledge

Gemini (Google)

Best for: Images, video analysis, multimodal applications

Choosing the Right Model

7LLM vs Deterministic

What is "Deterministic"?

Deterministic means: same input → same output, every time.

# Deterministic: 2 + 2 is always 4
def add(a, b):
    return a + b

add(2, 2)  # Always returns 4

When to Use What

Use Deterministic Code When:

• Calculate a total → Deterministic
• Validate an email format → Deterministic
• Convert CSV to JSON → Deterministic
• Apply discount if order > $100 → Deterministic

Use LLMs When:

• Summarize customer feedback → LLM
• Write a personalized email → LLM
• Determine if a review is positive or negative → LLM
• Answer customer questions → LLM

The Cost Reality

If you're processing 10,000 items:

• Deterministic: Instant, free
• LLM (Sonnet): 8+ hours, $100+

8Your First Project: A Simple Chatbot

What We're Building

A web-based chatbot that:

1. Shows a chat interface
2. Sends messages to Claude
3. Displays responses
4. Remembers conversation history
5. Saves conversations to a database
6. Runs 24/7 on the internet

Step 1: Project Setup

mkdir simple-chatbot
cd simple-chatbot
python -m venv .venv
source .venv/bin/activate
claude

Prompt to Claude Code:

"Create the basic project structure for a Streamlit chatbot app. I need requirements.txt with streamlit, anthropic, supabase, python-dotenv; .env.example; .gitignore; and an empty app.py file."

Step 2: Basic Streamlit Interface

Prompt to Claude Code:

"In app.py, create a simple Streamlit chat interface with a title, chat input, and message display. For now, just echo back what the user types."

Test it: streamlit run app.py

Step 3: Connect to Claude API

Prompt to Claude Code:

"Now connect the chatbot to Claude. Send user messages to the Claude API and display Claude's response."

Step 4: Add Conversation History

Prompt to Claude Code:

"Add conversation history so Claude remembers previous messages in the session using Streamlit session state."

Step 5: Save to Supabase

First, create a table in Supabase SQL Editor:

CREATE TABLE conversations (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    session_id TEXT NOT NULL,
    role TEXT NOT NULL,
    content TEXT NOT NULL,
    created_at TIMESTAMPTZ DEFAULT NOW()
);

Prompt to Claude Code:

"Add Supabase integration to save conversations. Generate a unique session ID and save each message."

Step 6: Deploy to Railway

Prompt to Claude Code:

"Create a Procfile and railway.json for Railway deployment."

Then:

1. Push to GitHub
2. Connect Railway to your repo
3. Add environment variables
4. Deploy!

9Common Pitfalls & How to Avoid Them

10Glossary

11What's Next?

You now have the foundation. You can set up a development environment, build simple AI applications, deploy them for others to use, and make informed decisions about when to use AI.

Keep Building

The best way to learn is to build things. Ideas to try:

• Modify your chatbot to focus on a specific topic
• Add a feature to export conversations
• Build a simple data dashboard
• Create a tool that summarizes documents