What is the difference between short-term and long-term memory in AI agents?

Short-term memory (also called working or in-context memory) is everything currently loaded into the LLM's context window — conversation history, tool results, and the current plan. It is fast but ephemeral: it vanishes when the session ends. Long-term memory is stored externally in a database or vector store, persists across sessions, and is retrieved on demand. Both are necessary in production agents.

What vector database should I use for my AI agent?

ChromaDB is the easiest starting point — it runs embedded in Python with no infrastructure. For production with managed SLAs, Pinecone is the most popular choice. If you already run PostgreSQL, the pgvector extension lets you skip a separate vector service entirely. The best choice depends on your existing stack, scale requirements, and tolerance for infrastructure management.

What is RAG and how does it relate to agent memory?

Retrieval-Augmented Generation (RAG) is the pattern of retrieving relevant documents from an external knowledge base and injecting them into the LLM context before generating a response. In the context of agent memory, RAG is the retrieval mechanism for semantic long-term memory — it's how an agent accesses its knowledge base without stuffing everything into the context window.

How do I prevent agent memory from consuming too many tokens?

Three strategies: (1) Cap short-term memory with a sliding window that discards old messages; (2) summarise older conversation history into a compact summary before the window; (3) for long-term memory, retrieve only the top-k most relevant chunks rather than injecting the full knowledge base. Set a hard max_memory_tokens budget and truncate retrieved content to fit. Memory cost control is as important as memory correctness.

What is episodic memory in AI agents and why does it matter?

Episodic memory records specific past interactions as retrievable episodes — what task was attempted, what actions were taken, and what outcome resulted. Unlike semantic memory (general knowledge), episodic memory is tied to specific events in time. It enables personalisation, audit trails for compliance, and the ability for agents to learn from past mistakes without model fine-tuning.

AI & Automation·9 min read·June 19, 2026

🧠 AI Agent Memory Systems: Short-Term, Long-Term, Episodic, and Vector Memory

AI agents need more than a reasoning engine — they need memory. This guide breaks down the four core memory types used in production agents, explains vector stores like ChromaDB, Pinecone, and pgvector, and shows Python patterns for managing context without exploding your token budget.

📖AI Roadmap for Electrical Engineers (Full Access)

View →

Why Memory Transforms an Agent from a Chatbot into a System

The single most important difference between a chatbot and a production AI agent is memory. Without memory, an agent resets on every request — it cannot track progress, recall prior decisions, or build on previous work. It degrades, as one practical guide on AI agent engineering puts it, into a reactive system rather than a goal-driven entity.

Memory is what lets an agent remember that three steps ago it called a web search tool, got three results, and is now on step four of an eight-step research plan. It is what lets a customer support agent recall that this user contacted support twice last month and escalated both times. It is what lets a coding assistant know which files it already analysed before suggesting a refactor.<

Topics covered

AI agent memoryshort-term memory agentlong-term memory AIvector databaseepisodic memory AIsemantic memoryChromaDBPineconepgvectorRAG pipelineretrieval-augmented generationconversation history managementsliding window memorymemory summarizationagent state managementLangChain memoryworking memory AIin-context memoryagent architectureLLM memory systems