26 Aug 2026

Memory Types in AI Agents

AI agent memory architectures rely on six distinct systems to manage context, persistent knowledge, and execution skills. Short-Term (Working Memory) holds temporary dialogue context within the active LLM window, while Long-Term Memory persists user state across separate sessions via external databases. Semantic Memory tracks discrete facts and user preferences, whereas Episodic Memory logs chronological event history so the agent can evaluate past outcomes and avoid repeating mistakes. Finally, Procedural Memory defines operational rules and workflow skills for executing multi-step tasks, and Vector/Retrieval Memory leverages dense embeddings to fetch relevant unstructured knowledge on demand via semantic similarity search.

  • AI
  • AI agent
  • Memory
Read more →

Ask about my work

Ask about my career, education, projects, or blog posts.