Core Concepts
Agent Memory in plain English.
Also known as: AI memory,long-term memory (AI),persistent memory
The one-sentence version
The mechanisms that let an AI agent remember facts, preferences, and past interactions across sessions rather than starting blank each time.
Agent memory is everything an AI system uses to carry information beyond a single conversation. A language model itself remembers nothing between requests; memory is built around it. Short-term memory is the context window — what is in front of the model right now. Long-term memory is stored outside the model and retrieved when relevant: facts about the user, decisions made in earlier sessions, documents, and the outcomes of past actions, usually kept in a database and pulled back in with retrieval techniques. Products like ChatGPT's memory feature and developer libraries such as Mem0 implement this. Good memory makes an assistant feel like it knows you; bad memory means stale facts resurfacing, private details remembered when they should not be, and rising costs as more history is stuffed into every prompt. Deciding what to remember, what to forget, and what to show the user is now a core design problem for agents.