---
title: "Prompt Caching · agent orchestration · Agent Fieldbook"
url: https://agentfieldbook.org/design/orchestration/prompt-cache/
description: "Cache the model's internal state for a long, stable prefix (system prompt, tool defs, docs) so repeat calls only pay for the new suffix. Vendor-native at the API layer."
section: "Design · Orchestration"
source: Agent Fieldbook — generated from the published page
---

# Prompt Caching

Cache the model's internal state for a long, stable prefix (system prompt, tool defs, docs) so repeat calls only pay for the new suffix. Vendor-native at the API layer.

Verified against [docs.anthropic.com](https://docs.anthropic.com/en/docs/build-with-claude/prompt-caching)

- [docs.anthropic.com](https://docs.anthropic.com/en/docs/build-with-claude/prompt-caching)

## How it works

## At a glance

| Field | Value |
| --- | --- |
| Technique | Caching |
| Cost lever | Lower, cached prefix tokens billed at ~10% of input rate |
| Latency lever | Lower, cached prefix skips recompute |
| Quality lever | Identical, same model, same context |
| Example | Anthropic prompt caching· OpenAI / Gemini cached content |

## When to use

Agents with large fixed context (long system prompt, big tool schema, RAG docs) reused across many turns.

## Basis

## Verified against

- [https://docs.anthropic.com/en/docs/build-with-claude/prompt-caching](https://docs.anthropic.com/en/docs/build-with-claude/prompt-caching)

The corpus does not rewrite vendor documentation. Where copy is quoted verbatim it is marked as such and attributed to its source.