⚡ Your AI bill is eating your Q4 margin.Protect Your Margin. Get a FREE 360° Cost ReportFree 360° Cost Report

Home › Docs › Caching

How it works

Caching

Caching is usually the biggest single saving for apps that resend long prompts. apicuts preserves and maximizes provider prompt, prefix, and context caching across every eligible request.

Provider prompt, prefix, and context caching

OpenAI, Anthropic, and Google bill repeated prompt prefixes and context blocks at sharply discounted cached rates (via automatic prefix caching and Anthropic cache_control blocks). apicuts keeps those discounts intact across all three gateway endpoints:

  • Static first, dynamic last. Put stable system instructions, tool schemas, and reference documents at the start of the prompt (for example, in the top-level system field on Anthropic /v1/messages or shared prefixes on /v1/responses) before per-turn user input.
  • Cache-aware routing. Requests that share a prefix stay routed to the provider holding that prefix cache so hit rates remain high instead of thrashing across suppliers.
  • Measured, not assumed. Your workspace ledger (/app/overview and /app/requests, or preview /demo/overview) separates cache read and cache write tokens on every request.

Background: prompt caching, explained.

What caching means for your data

apicuts does not store prompts or model outputs in a gateway response cache. Provider prompt and context caches live only inside the upstream provider for that provider's short ephemeral cache window. See security and privacy.

Want us to find your 20%+?

Get a free 360° cost report. Bring last month's provider usage and we'll show you your biggest waste and the first fix we'd ship.

Get a Free 360° Cost Report