<Section title="Actuators in the agent anatomy">## What actuators are in this material In Silvio Savarese's decomposition of an agent, actuators are the components that make API and function calls — the part of the agent that reaches outside the model and actually does something. He splits an agent into four parts: memory (typically RAG), a brain or reasoning engine, actuators, and an interface. [^50a3b3fa-9ad2-59a3-9c76-ddfe74635066@220] <Quote moment="220">actuators (API and function calls)</Quote> ### Why they matter Savarese argues that an agent without connected data and context is useless — the actuators are precisely what connects it. Enterprise deployments stumble here: when data is siloed (sales and marketing manage data differently), the agent's ability to act on the world breaks down because the dots aren't connected. [^50a3b3fa-9ad2-59a3-9c76-ddfe74635066@220] [^50a3b3fa-9ad2-59a3-9c76-ddfe74635066@342] ### Where they fit in the bigger picture Across the channel's material, actuation is the reason agents need infrastructure at all: for agents to be useful they must take action on your behalf, plumb into existing APIs, and operate through protocols like MCP — all of which runs through an agent's actuators. </Section>