Using the API

• Use the API when you need to call an LLM from Python code inside a notebook, a script, or a pipeline
• Authentication is done via an environment variable, not hardcoded in source files

export ANTHROPIC_API_KEY="sk-ant-..."   # add to ~/.zshrc or ~/.bashrc

import anthropic
client = anthropic.Anthropic()   # reads ANTHROPIC_API_KEY automatically
response = client.messages.create(
    model="claude-opus-4-6",
    max_tokens=512,
    messages=[{"role": "user", "content": "How many rows are in a 10x10 grid?"}]
)
print(response.content[0].text)

• The response object contains content (a list of content blocks), usage (token counts), and stop_reason
• Cost is (input_tokens * input_price) + (output_tokens * output_price)
  • Check the provider's pricing page
• Rate limits apply: requests per minute and tokens per minute
  • Exceeded limits raise HTTP 429 errors
• Use exponential back-off when retrying rate-limited requests

import time
for attempt in range(5):
    try:
        response = client.messages.create(...)
        break
    except anthropic.RateLimitError:
        time.sleep(2 ** attempt)

Exercises

• Call the LLM API from a short Python script to summarize a dataset
  • Print the usage field of the response
  • Record how many input and output tokens the call consumed and estimate its cost