Lesson 06 — How to use an API

From typing once to running a thousand times — while you make coffee.

AI for Dummies Module 6

By the end of this lesson you’ll have run a real API call, understood what a token is, and know how to keep costs under control.

Why an API? — 1 vs 1,000

The kitchen window — same chef, different door

Same restaurant. Same chef. Same food. Just a different, programmable doorway.

Anatomy of an API call

Every language — curl, Python, JavaScript — wraps this same HTTP round-trip. The shape never changes.

Step 1 — Get an API key

Your API key — treat it like a credit card number

The simplest call — curl in the terminal

No library to install. curl is on every Mac and Linux machine already.

curl https://api.anthropic.com/v1/messages \        # ← the endpoint
  -H "x-api-key: $ANTHROPIC_API_KEY" \              # ← key from env var, never in plain text
  -H "anthropic-version: 2023-06-01" \              # ← API version header (required)
  -H "content-type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",                   # ← which model to use
    "max_tokens": 1024,                              # ← ceiling on reply length
    "messages": [
      {"role": "user", "content": "Explain what an API is in one sentence."}
    ]
  }'

Same call in Python — a complete AI program

Install once: pip install anthropic. Then this is a complete working program.

import anthropic

client = anthropic.Anthropic()   # reads ANTHROPIC_API_KEY from environment automatically

message = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    messages=[
        {"role": "user", "content": "Write a friendly 2-line welcome email."}
    ],
)

print(message.content[0].text)   # ← the model's reply, as a plain string

What just happened? — the round trip

Total round-trip for a short reply: typically 1–3 seconds. For streaming responses you can receive tokens as they're generated — but that’s a later lesson.

Local model — same shape, free & private

import anthropic

client = anthropic.Anthropic()

msg = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=256,
    messages=[{"role":"user",
               "content":"Hello!"}],
)
print(msg.content[0].text)

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:11434/v1",
    api_key="ollama",  # any string
)
resp = client.chat.completions.create(
    model="llama3.2",
    messages=[{"role":"user",
               "content":"Hello!"}],
)
print(resp.choices[0].message.content)

Ollama vs LM Studio — two local servers

base_url="http://localhost:11434/v1"
model="llama3.2"  # or gemma3, mistral…

base_url="http://localhost:1234/v1"
model="lmstudio-community/Meta-Llama-3…"

Covered in depth in Lesson 04. Revisiting it here because it unlocks free, private API calls for any volume.

What is a token?

Cloud APIs don’t charge per message or per word — they charge per token. Here’s what that means in practice:

Model pricing tiers — tiny to powerful

Prices are per million tokens. Same idea, very different scale.

Input / output rates. Check provider docs for current prices.

Set a spending limit — do this before anything else

# Always set max_tokens explicitly!
# Without it, the model can return
# thousands of tokens per call.
message = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=256,      # ← hard ceiling
    messages=[...]
)

Local models — the cost is always $0

Smart path: prototype with the cloud (no setup), switch to local once cost or privacy is the bottleneck. Because both use the same API format, the switch is one URL change.

Realistic example — classify a 1,000-row spreadsheet

This is the most common beginner use-case. The loop is the whole idea.

import anthropic, csv

client = anthropic.Anthropic()
results = []

with open("reviews.csv") as f:
    for row in csv.DictReader(f):               # ← iterate each row
        msg = client.messages.create(
            model="claude-haiku-4-5",           # ← cheapest model for bulk work
            max_tokens=20,                       # ← short reply: just "positive" / "negative"
            messages=[{"role": "user", "content":
                f"Classify as positive/negative: {row['review_text']}"
            }],
        )
        results.append({
            "id":      row["id"],
            "label":   msg.content[0].text.strip(),
        })

print(f"Done. {len(results)} rows classified.")

Recap — what you now know