LangMem

Overview

LangMem is LangChain's official library for adding long-term memory to LangGraph agents. It provides memory extraction, storage, and retrieval capabilities designed to integrate seamlessly with the LangChain ecosystem.

Key Features

**Memory Extraction**: Automatically extracts memorable information from conversations

**Semantic Memory**: Store and query facts and preferences

**LangGraph Integration**: Native integration with LangGraph agents

**Flexible Storage**: Supports various vector store backends

**Memory Namespaces**: Organize memories by user, session, or custom scopes

When to Use LangMem

LangMem is ideal for:

Existing LangChain/LangGraph projects needing memory

Teams already invested in the LangChain ecosystem

Applications requiring tight integration with LangGraph workflows

Building agents that learn from conversations

Pros

Official LangChain support

Seamless LangGraph integration

Flexible and extensible

Good documentation

Active development

Cons

Requires LangChain/LangGraph knowledge

Relatively new library

Less standalone than other options

Tied to LangChain ecosystem

Getting Started

from langmem import create_memory_store
from langgraph.prebuilt import create_react_agent

# Create memory store
memory = create_memory_store(
    embeddings=embeddings,
    store=vector_store
)

# Use with LangGraph agent
agent = create_react_agent(
    model=model,
    tools=tools,
    memory=memory
)

Architecture

LangMem uses a modular architecture:

**Memory Store**: Handles storage and retrieval

**Memory Extractors**: Pull memorable info from conversations

**Memory Injectors**: Add relevant memories to prompts

Pricing

**Open Source**: Free, Apache 2.0 license

Memory Types

Integrations