video1.0<iframe src="https://www.loom.com/embed/48222cbddadc4d12829879dea86ed512" frameborder="0" width="3376" height="2532" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>25323376Loomhttps://www.loom.com25323376https://cdn.loom.com/sessions/thumbnails/48222cbddadc4d12829879dea86ed512-2a83a699dafd4dbd.gif33.033333Prompt caching and fallback connections help Make AI Agents run more reliably and efficiently. Prompt caching lets supported models reuse stable parts of an agent’s setup, such as instructions and tool definitions, so repeat calls can respond faster and use fewer input tokens. Fallback connections give an agent a backup path when its primary LLM connection is unavailable, automatically retrying the run once with the configured fallback. Together, they help your agents stay efficient during repeat work and keep important automations moving through outages, timeouts, rate limits, or expired connections.