Context Profiles

Context Profiles optimize LLM efficiency by caching massive datasets (up to 100k+ tokens) for instant, low-cost retrieval across multiple API calls.

Context Profiles eliminate the latency and cost of re-processing static data. By snapshotting large-scale documentation, codebases, or legal archives (think 200-page technical manuals), developers can inject deep domain knowledge into every prompt without paying the full token price each time. This architecture reduces time-to-first-token by up to 90% for high-context applications. It is the standard for building responsive RAG systems and AI agents that require persistent, high-fidelity memory of specific project environments.

https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/context-caching