★ OverviewIntermediate11 min

Agent Memory Systems

Short-term, long-term, and episodic memory for agents — how to make your agent remember context across conversations and learn from experience.

Quick Reference

→Short-term memory = conversation history in the graph state (messages list), lost when the thread ends
→Long-term memory = persisted key-value store (LangGraph Store API) that survives across threads
→Episodic memory = summaries of past interactions stored as searchable documents for retrieval
→Use MemorySaver for checkpointing thread state; use Store for cross-thread user preferences and facts
→Token management: summarize or trim old messages to stay within the context window budget

The Three Memory Types

Three tiers of agent memory: ephemeral buffers, cross-session stores, and permanent episodic recall

Type	Scope	Persistence	Implementation	Example
Short-term	Current conversation	Thread lifetime only	MessagesState (messages list)	Last 10 messages in this chat
Long-term	Across all conversations	Permanent (Store API)	store.put() / store.search()	User prefers dark mode, speaks Spanish
Episodic	Across conversations	Permanent (vectorized)	Embedded summaries in vector store	Last week user asked about refund policy

Short-Term Memory

Short-term memory is the conversation history — the messages list in your graph state. Every message the user sends and every response the agent generates is stored here. It's automatically managed by LangGraph's MessagesState.

Long-Term Memory with the Store API

Long-term memory survives across conversations. Use the LangGraph Store API to save and retrieve user preferences, facts, and instructions that should persist:

Sign in to read this article

This is a premium article. Sign in with your Google account to continue.