Memory agents persist context — goals, facts, or user preferences — across sessions. This persistence allows continuity of reasoning and personalization. Frameworks like MemGPT or LangGraph Memory show how long-term memory transforms reactive LLMs into adaptive systems.