Agent Memory

Agent memory is the mechanism by which an AI agent retains information across steps or sessions — such as past actions, facts, or user preferences — beyond what fits in a single context window.

Because a context window is finite, agents store and retrieve relevant history externally so they can maintain continuity over long or repeated tasks.

Memory is commonly split into short-term working context and longer-term stores like episodic and semantic memory, often backed by a vector database.