How to Use Prompt Caching to Save Money on AI

# Python example - Anthropic API with Prompt Caching
import os
import anthropic

client = anthropic.Anthropic(api_key=os.environ.get("ANTHROPIC_API_KEY"))

# This system prompt gets cached after first request
system_prompt = """
You are a luxury real estate copywriter specializing in Nashville, TN.
Market context: Median $445K, 2.8 months inventory, 24 avg DOM.

Style guidelines:
- Short, punchy sentences
- Lead with strongest visual feature
- Under 150 words
- End with location context
- Never use: stunning, nestled, welcome to, boasts

Examples of my voice:
[3 example listing descriptions would go here]
"""

# Process a batch of listings with the same cached system prompt
properties = [
    "4BR/3BA colonial, Franklin, 3200 sqft, pool, renovated kitchen",
    "2BR/2BA condo, Germantown, 1100 sqft, rooftop access, modern finishes",
    "3BR/2BA ranch, Donelson, 1800 sqft, hardwood floors, fenced yard",
]

for prop in properties:
    message = client.messages.create(
        model="claude-sonnet-4-5-20250929",
        max_tokens=300,
        system=[
            {
                "type": "text",
                "text": system_prompt,
                "cache_control": {"type": "ephemeral"}
            }
        ],
        messages=[{"role": "user", "content": f"Write listing description: {prop}"}]
    )
    
    # Check cache usage
    print(f"Cache read tokens: {message.usage.cache_read_input_tokens}")
    print(f"Cache creation tokens: {message.usage.cache_creation_input_tokens}")
    print(message.content[0].text)
    print("---")