Mission Control · API bay · §4.1

Prompt Caching Optimizer

Paste a prompt and we'll auto-segment it, classify each section stable / volatile, suggest the optimal cache_control breakpoint placement, and show the monthly savings across Anthropic, OpenAI, and Gemini.

Commander tier required

Locked on the recon tier.

Prompt Caching Optimizer is part of the API bay. Commander unlocks all 6 API-bay tiles + all 8 Code-bay tiles, $25 AI budget, public REST API, and CLI plugin for $24/mo.

See tiers

How the optimizer thinks

Cache works on prefixes, not on chunks. If a single volatile block sits before your stable reference docs, NONE of those docs are cacheable. Stable-first ordering is the whole game.

Auto-segmentation honors structure. The optimizer prefers XML tags (<context>, <example>), markdown headings, and role markers (User:, Context:) as section boundaries.

Anthropic minimum is ~1k–2k tokens. Cache_control on a short prompt is a no-op. The optimizer flags this — don't pay for the cache-write overhead if you're under the floor.

Up to 4 breakpoints. Anthropic accepts 4 cache_control breakpoints per request. The optimizer suggests the primary one (after the last stable block) plus secondaries when stable content is unusually large.