Mastering Cache Hits in Claude Code

Understanding how caching works behind the scenes so you can reduce costs and get faster responses — even though you never touch the API directly. Table of Contents What Are Cache Hits and Why Should I Care? Anatomy of an API Call Cache Hits and Misses Explained What Breaks the Cache Cache Lifetime and the TTL Timer Structuring Your Work for Better Caching Caching Anti-Patterns API-Level Details (For When You Need Them) References What Are Cache Hits and Why Should I Care? Every time Claude Code sends a message on your behalf, it makes an API call to Anthropic. That API call includes everything Claude needs to respond: the system prompt, any tool definitions, your CLAUDE.md files, and your entire conversation history. On a long session with a big codebase loaded, this can easily be 50,000–200,000+ tokens of input. Without caching, Anthropic's servers have to fully process all of those tokens from scratch on every single message — even though 99% of them are identical to what was sent 3