Tutorials

Claude API: Getting Started Guide (With Practical Examples)

April 19, 2024 4 min read Updated: 2026-02-03

Claude API: Getting Started Guide

I’ve used the Claude API for various projects over the past year. Here’s the practical guide I wish I had when starting.

Getting Access

  1. Go to console.anthropic.com
  2. Create account or sign in
  3. Add payment method (can use $5 free credits first)
  4. Generate API key

No waitlist. Instant access now.

The Models

As of early 2026:

ModelBest ForInput CostOutput Cost
Claude 3.5 SonnetBest overall$3/M tokens$15/M tokens
Claude 3 OpusComplex reasoning$15/M tokens$75/M tokens
Claude 3.5 HaikuFast, cheap$0.25/M tokens$1.25/M tokens

My recommendation: Start with Sonnet. Use Haiku for high-volume simple tasks. Use Opus only when you need maximum capability.

Basic API Call (Python)

import anthropic

client = anthropic.Anthropic(api_key="your-key-here")

message = client.messages.create(
    model="claude-3-5-sonnet-20241022",
    max_tokens=1024,
    messages=[
        {"role": "user", "content": "Explain recursion simply."}
    ]
)

print(message.content[0].text)

That’s it. The simplest possible call.

Adding System Prompts

System prompts set behavior and context:

message = client.messages.create(
    model="claude-3-5-sonnet-20241022",
    max_tokens=1024,
    system="You are a senior Python developer. Give concise, practical advice.",
    messages=[
        {"role": "user", "content": "How should I structure a Flask app?"}
    ]
)

System prompts are powerful. Use them to:

  • Set tone and expertise level
  • Define output format
  • Establish constraints

Multi-Turn Conversations

messages = [
    {"role": "user", "content": "I'm building a todo app in Python."},
    {"role": "assistant", "content": "Great! What framework are you considering?"},
    {"role": "user", "content": "Flask. How should I structure the database?"}
]

message = client.messages.create(
    model="claude-3-5-sonnet-20241022",
    max_tokens=1024,
    messages=messages
)

Pass the full conversation history. Claude uses it for context.

Streaming Responses

For real-time output (chatbots, etc.):

with client.messages.stream(
    model="claude-3-5-sonnet-20241022",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Write a short story."}]
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

Streaming makes applications feel responsive.

Handling Long Documents

Claude’s context window is huge (200K tokens). Use it:

# Reading a long document
with open("long_document.txt", "r") as f:
    document = f.read()

message = client.messages.create(
    model="claude-3-5-sonnet-20241022",
    max_tokens=4096,
    messages=[
        {"role": "user", "content": f"Summarize this document:\n\n{document}"}
    ]
)

Claude handles long contexts better than most competitors.

Practical Examples

Code Review

code = """
def get_user(id):
    user = db.query(f"SELECT * FROM users WHERE id = {id}")
    return user
"""

message = client.messages.create(
    model="claude-3-5-sonnet-20241022",
    max_tokens=1024,
    system="You are a security-focused code reviewer.",
    messages=[
        {"role": "user", "content": f"Review this code for issues:\n\n{code}"}
    ]
)

Structured Output (JSON)

message = client.messages.create(
    model="claude-3-5-sonnet-20241022",
    max_tokens=1024,
    system="Always respond in valid JSON.",
    messages=[
        {"role": "user", "content": "Extract entities from: 'Apple announced iPhone 15 in Cupertino.' Return JSON with company, product, location."}
    ]
)

Content Generation

message = client.messages.create(
    model="claude-3-5-sonnet-20241022",
    max_tokens=2048,
    system="You are a professional copywriter. Write engaging, clear content.",
    messages=[
        {"role": "user", "content": "Write product description for wireless earbuds. Features: 30hr battery, ANC, comfortable fit. Target: busy professionals."}
    ]
)

Cost Optimization Tips

1. Use the right model

  • Haiku for simple tasks (classification, extraction)
  • Sonnet for most use cases
  • Opus only when necessary

2. Limit output tokens

max_tokens=500  # Don't set higher than needed

3. Be concise in prompts

More input tokens = more cost. Keep prompts focused.

4. Cache responses

Don’t re-call API for identical queries.

5. Use prompt caching

For repeated system prompts, Anthropic offers prompt caching to reduce costs.

Common Mistakes

1. Not handling errors

try:
    message = client.messages.create(...)
except anthropic.APIError as e:
    print(f"API error: {e}")
except anthropic.RateLimitError:
    print("Rate limited, backing off...")
    time.sleep(60)

Always handle errors. API calls fail sometimes.

2. Ignoring rate limits

Don’t spam requests. Implement backoff:

import time

def call_with_retry(prompt, max_retries=3):
    for attempt in range(max_retries):
        try:
            return client.messages.create(...)
        except anthropic.RateLimitError:
            wait = 2 ** attempt
            time.sleep(wait)
    raise Exception("Max retries exceeded")

3. Hardcoding API keys

Use environment variables:

import os
client = anthropic.Anthropic(api_key=os.environ["ANTHROPIC_API_KEY"])

Never commit API keys to code.

Monitoring Usage

Track your costs at console.anthropic.com:

  • Usage dashboard shows token consumption
  • Set up billing alerts
  • Monitor by project if using workspace features

When to Use Claude API vs Chat

Use the API when:

  • Building applications
  • Need programmatic access
  • Processing data at scale
  • Integrating with other systems

Use Claude chat (claude.ai) when:

  • Interactive work
  • One-off tasks
  • Exploration and brainstorming

Quick Reference

Endpoint: https://api.anthropic.com/v1/messages

Headers:

  • x-api-key: YOUR_KEY
  • anthropic-version: 2023-06-01
  • content-type: application/json

Best model for most cases: claude-3-5-sonnet-20241022

Max context: 200,000 tokens

Max output: 8,192 tokens per response

Bottom Line

The Claude API is straightforward to use. The hardest part is getting your prompts right, not the technical integration.

Start with Sonnet for most use cases. Use system prompts to shape behavior. Handle errors properly. Monitor costs.

The rest is just application logic.

Frequently Asked Questions

Claude 3.5 Sonnet costs $3 per million input tokens, $15 per million output tokens. Claude 3 Opus is more expensive at $15/$75. For most projects, Sonnet offers the best value. Typical use costs $5-50/month for moderate usage.

For long-form writing and analysis, Claude is often better. For coding and quick tasks, OpenAI is competitive. Claude handles longer contexts better (200K tokens). Best choice depends on your use case.

Sign up at console.anthropic.com. Add payment method. Get API key. No waitlist anymore - instant access. Free trial credits available to start.

Disclosure: This post contains affiliate links. If you click through and make a purchase, we may earn a commission at no extra cost to you. We only recommend tools we genuinely believe in.

Related Articles