Spring AI Prompt Caching: Stop Wasting Money on Repeated Tokens

Prompt caching is a technique to reduce LLM API costs by caching static parts of prompts like system messages and tool definitions, so they aren't re-processed on every request. A Spring AI application is built using the Anthropic Claude integration, demonstrating how to configure AnthropicChatOptions with a caching strategy (system-only) to cache a long system prompt. On the first request, cache creation tokens are logged; on subsequent requests, cache read tokens replace the full input cost, yielding up to 90% savings on cached tokens with models like Claude Sonnet 4.5.