Your First AI App in Code

In Chapter 6 your chatbot followed rules you wrote. Now you'll plug in a real AI brain — Claude — so your program can answer almost anything. This is the moment your code becomes genuinely intelligent.

How code talks to an AI

Your program sends a prompt over the internet to the AI company's computers (an API — Application Programming Interface), and gets the reply back. You need three things:

1. The anthropic Python package (the toolkit for talking to Claude).
2. An API key — a secret password that says "this is my account."
3. A few lines of code.

Step 1 — Install the toolkit

In a new Colab cell (from Chapter 6), run:

!pip install anthropic

The ! tells Colab to install a package. You only do this once per notebook.

Step 2 — Connect with your key

This code asks for your key without showing it on screen, so it stays secret:

from anthropic import Anthropic
from getpass import getpass

api_key = getpass("Paste your Anthropic API key (it stays hidden): ")
client = Anthropic(api_key=api_key)
print("Connected! ✅")

client is now your line to Claude.

Step 3 — Your first AI reply in code

message = client.messages.create(
    model="claude-haiku-4-5",
    max_tokens=300,
    system="You are a friendly study buddy for a curious kid. Keep answers short and clear.",
    messages=[
        {"role": "user", "content": "Explain what gravity is in 2 sentences."}
    ],
)

print(message.content[0].text)

Look closely — you already know these ideas from Chapter 3!

• model — which AI to use. claude-haiku-4-5 is fast and low-cost: perfect for projects.
• system — the persona and rules (your "standing orders").
• messages — the conversation. Each message has a role (user or assistant) and content.
• max_tokens — a limit on how long the reply can be (keeps it short + cheap).

message.content[0].text pulls the words out of the reply.

Step 4 — Turn it into a reusable helper

Wrap it in a function so you can ask anything with one line:

def ask_buddy(question):
    message = client.messages.create(
        model="claude-haiku-4-5",
        max_tokens=300,
        system="You are a friendly study buddy for a curious kid. Keep answers short and clear.",
        messages=[{"role": "user", "content": question}],
    )
    return message.content[0].text

print(ask_buddy("Why is the sky blue?"))
print(ask_buddy("Give me one cool fact about octopuses."))

Project — Your command-line study buddy 🛠️

Let's make a program you can actually chat with. It keeps asking questions until you type "bye":

print("🤖 Study Buddy is ready! Ask me anything. Type 'bye' to stop.\n")

while True:
    question = input("You: ")
    if question.lower() == "bye":
        print("Buddy: See you later! 👋")
        break
    answer = ask_buddy(question)
    print("Buddy:", answer, "\n")

1. Run it and have a real conversation.
2. Make it yours: change the system prompt to create a different helper — a joke-teller, a Spanish tutor, a riddle master.
3. Add a touch: print "Buddy is thinking..." before the answer so it feels alive.

try:
    answer = ask_buddy(question)
except Exception as e:
    answer = "Oops, something went wrong. Try again!"
    print("(debug:", e, ")")

What you learned

• Code talks to Claude through an API using a secret key.
• client.messages.create(...) sends a system prompt + messages and returns a reply.
• Your prompt-engineering skills work the same in code.
• You built a real, interactive AI helper. 🎉

You've earned the API Apprentice badge. 🏅

Your helper answers questions — but it forgets everything between them, and it can't do anything in the world. Next you'll fix both: memory and tools. That's Chapter 8: Giving AI Memory & Tools.