# LangMem

> Open-source SDK from LangChain for long-term memory in LLM agents, with hot-path tools, a background memory manager, and native LangGraph storage integration.

LangMem is a Python SDK for adding long-term memory to LLM agents. It provides two complementary paths:

- Hot path tools (create_manage_memory_tool, create_search_memory_tool) that agents call during conversations to store and retrieve facts, preferences, and other context.
- Background manager that periodically extracts, consolidates, and updates memories outside the request path.

The core API is storage-agnostic and works with LangGraph’s BaseStore implementations (e.g., in-memory for dev, Postgres for production). It supports semantic and episodic memory patterns, prompt optimization/“procedural memory,” and dynamic namespaces for user- or team-scoped memories. Typical setups pair LangMem with LangGraph agents, Postgres/pgvector for persistence, and common model providers (Anthropic, OpenAI).

Install: pip install -U langmem 
Requires: Python 3.10+

Use it to build agents that remember user preferences across sessions, keep evolving profiles, and refine prompts over time—without hand-rolling memory extraction, deduplication, and retrieval plumbing.

## Features
- Hot-path memory tools for store/search within agent runs
- Background memory manager for batch extraction and consolidation
- Storage-agnostic core API; works with LangGraph BaseStore
- Native integrations for InMemoryStore and AsyncPostgresStore
- Dynamic namespaces for user/team-scoped memories
- Supports semantic, episodic, and procedural (prompt) memory patterns
- Embeddings-based retrieval (e.g., OpenAI text-embedding-3-small)
- Typed/structured memories via Pydantic models
- CrewAI and custom-agent usage guides
- MIT-licensed; Python package via PyPI

## Integrations
LangGraph, LangGraph Platform, PostgreSQL (pgvector), Anthropic, OpenAI, CrewAI

## Platforms
DEVELOPER_SDK, WEB, WINDOWS, MACOS, LINUX

## Pricing
Open Source, Free tier available

## Version
0.0.29

## Links
- Website: https://langchain-ai.github.io/langmem/
- Documentation: https://langchain-ai.github.io/langmem/
- Repository: https://github.com/langchain-ai/langmem/
- EveryDev.ai: https://www.everydev.ai/tools/langmem