Build Your First AI Chatbot with the Anthropic Claude SDK (2026)

This tutorial walks you through building your first AI chatbot with the Anthropic Claude SDK. By the end you will have a command-line chatbot that uses Claude Sonnet 5.5, remembers the conversation, streams responses token by token, and handles common errors cleanly. The whole thing runs in about 60 lines of code and takes 20 minutes.

What is the Anthropic Claude SDK

The Anthropic Python SDK is the official library for calling Claude models from your code. In 2026 the SDK supports every Claude model (Opus 5.5, Sonnet 5.5, Haiku 4.5), streaming, tool use, prompt caching, and the Messages API’s full feature set. The SDK installs in one command and the API shape is clean and consistent across the model family.

Why Claude for a first chatbot: Claude’s instruction-following is excellent out of the box, especially on long or nuanced system prompts. For first projects where you want the model to stay on-persona and follow your rules, Claude is forgiving.

Prerequisites

• Python 3.10 or newer
• An Anthropic account (sign up at console.anthropic.com)
• API credit on your account ($5 minimum to activate; you will spend under $1 for this tutorial)
• Your API key, which starts with sk-ant-

Step 1: Install and set up

mkdir claude-chatbot && cd claude-chatbot
python -m venv .venv
source .venv/bin/activate
pip install anthropic python-dotenv

Create a .env file:

ANTHROPIC_API_KEY=sk-ant-...

Step 2: Your first Claude call

Create hello.py:

from anthropic import Anthropic
from dotenv import load_dotenv

load_dotenv()
client = Anthropic()

response = client.messages.create(
    model="claude-sonnet-5-5",
    max_tokens=512,
    messages=[{"role": "user", "content": "Explain in two sentences what Claude is."}],
)
print(response.content[0].text)

Run python hello.py. You should see a two-sentence explanation.

Three things to note:

• max_tokens is required on every call (unlike OpenAI where it is optional)
• The response’s content is a list of blocks, not a single string. For text responses, grab content[0].text.
• Model names use the hyphenated format (claude-sonnet-5-5, claude-opus-5-5, claude-haiku-4-5-20251001)

Step 3: Add a system prompt

Unlike OpenAI, Anthropic takes the system prompt as a top-level parameter, not a message in the list:

response = client.messages.create(
    model="claude-sonnet-5-5",
    max_tokens=512,
    system="You are a helpful Python tutor. Keep answers under 100 words and always include a code example when relevant.",
    messages=[{"role": "user", "content": "What is a dictionary?"}],
)
print(response.content[0].text)

The system field is where Claude’s instruction-following shines. Detailed system prompts (multi-paragraph persona, rules, examples) consistently shape output quality.

Step 4: Keep the conversation going

For a multi-turn chat, append each turn to the messages list. The system prompt stays separate.

messages = []
system = "You are a helpful Python tutor. Keep answers under 100 words."

for user_input in ["What is a dictionary?", "Can you show me how to iterate over one?"]:
    messages.append({"role": "user", "content": user_input})
    response = client.messages.create(
        model="claude-sonnet-5-5",
        max_tokens=512,
        system=system,
        messages=messages,
    )
    reply = response.content[0].text
    print(f"\nYou: {user_input}\nClaude: {reply}")
    messages.append({"role": "assistant", "content": reply})

The second question (“Can you show me how to iterate over one?”) makes sense because the full history is in the list. Claude knows “one” refers to a dictionary.

Step 5: Stream the response

Streaming makes the chat feel responsive even on longer answers. Claude’s streaming API uses a context manager:

with client.messages.stream(
    model="claude-sonnet-5-5",
    max_tokens=512,
    system="You are a helpful assistant.",
    messages=[{"role": "user", "content": "Describe the game of chess in a short paragraph."}],
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)
    print()

The stream.text_stream iterator yields string chunks as they arrive. For more granular control (events, token counts, errors), iterate over the raw stream events instead.

Step 6: Build the full chatbot

Put it together as a conversation loop with streaming and basic error handling:

from anthropic import Anthropic, APIError
from dotenv import load_dotenv

load_dotenv()
client = Anthropic()

system = (
    "You are a friendly assistant. Keep answers under 150 words. "
    "If you do not know the answer, say so instead of guessing."
)
messages = []

print("Chat with Claude. Type 'quit' to exit.")
while True:
    user = input("\nYou: ").strip()
    if user.lower() in ("quit", "exit", ""):
        print("Bye!")
        break

    messages.append({"role": "user", "content": user})
    try:
        print("\nClaude: ", end="", flush=True)
        full = ""
        with client.messages.stream(
            model="claude-sonnet-5-5",
            max_tokens=512,
            system=system,
            messages=messages,
        ) as stream:
            for text in stream.text_stream:
                print(text, end="", flush=True)
                full += text
        print()
        messages.append({"role": "assistant", "content": full})
    except APIError as e:
        print(f"\n(API error: {e}. Try again.)")
        messages.pop()  # remove the user message so we can retry cleanly

Costs and model picks

Claude pricing in September 2026 (per million tokens):

• Claude Haiku 4.5: $1 input / $5 output (fastest, cheapest)
• Claude Sonnet 5.5: $3 input / $15 output (production default)
• Claude Opus 5.5: $15 input / $75 output (top-tier reasoning)

For a learning chatbot, Haiku 4.5 handles everything you will throw at it for cents. For most production chat applications, Sonnet 5.5 is the right default.

Frequently Asked Questions

Angel Jude Suarez

Full-Stack Developer at PIES IT Solution

Focuses on Python development, machine learning, and AI integration. Has built production AI systems including OpenAI Whisper integration for medical transcription and GPT-4o-powered diagnosis assistance. Strong background in pandas, scikit-learn, and TensorFlow.

Expertise: Python, PHP, Java, VB.NET, ASP.NET, Machine Learning, AI Integration, OpenCV, Django, CodeIgniter · View all posts by Angel Jude Suarez →