Prompt caching reduces costs for AI agents on Amazon Bedrock

Amazon Bedrock introduces prompt caching to lower costs for AI agents. The service stores the input prompt, including tools and instructions, and reuses the processed state on subsequent requests. This avoids re-processing the entire context, reducing input token prices. Caching can be implicit, where Bedrock automatically identifies reusable prefixes, or explicit, where developers specify the cache key. The feature is particularly beneficial for multi-turn conversations and agents that spawn helpers, as it maintains consistency while cutting expenses.