Cutting costs with Claude prompt caching
If you send the same large context repeatedly — system prompts, files, tool definitions — caching turns those tokens from expensive to nearly free.
·
How caching saves money
Cache writes and cache reads are metered separately, and cache reads cost a fraction of fresh input tokens. Stable, reused context is the ideal candidate.
See also: How to save tokens on the Claude API
It stacks with your discount
Caching lowers the token count; your apiToken.sale discount lowers the price per token. Together they compound into a much smaller bill, and every cache line is visible in your usage breakdown.
Frequently asked questions
How much does prompt caching save?
Cache reads cost a fraction of fresh input tokens, so repeated large context becomes far cheaper.
Does caching work with the discount?
Yes — caching reduces token count and the discount reduces price per token, so the savings multiply.
Start with Google or GitHub and get $10 of Claude usage at official prices — no card required.