LangMem
LangChainLangMem is a Python library that lets agents extract, consolidate and search long-term memories from their conversations, either through memory tools the agent calls during a chat or through a background manager that runs afterward. Its memory types include semantic facts and user profiles, episodic examples and procedural memory, where it rewrites the agent's own prompt. LangChain builds it, with native support for LangGraph's memory store.
Openness
5 high confidence- license
- MIT (langmem library)
- source
- public
- core features withheld
- no — the 91-path tree at 4beb9c3 holds one MIT LICENSE and no ee/, enterprise/, commercial/ or proprietary/ path, and its dependencies are LangChain's open libraries; LangChain's paid products are the LangSmith platform
- commercial
- LangSmith
LangMem is a small MIT-licensed Python library, and all of it is in the public repository. Its memory primitives work with any storage system, including LangGraph's open in-memory and Postgres stores, and LangChain sells its hosted LangSmith platform as a separate product.
- https://raw.githubusercontent.com/langchain-ai/langmem/main/LICENSE recorded 2026-11-20
The root LICENSE is the MIT License with "Copyright (c) 2025 LangChain".
- https://raw.githubusercontent.com/langchain-ai/langmem/main/pyproject.toml recorded 2026-11-20
The manifest names the package langmem, takes its license from the LICENSE file, and depends only on langchain, langgraph, langsmith, trustcall and provider packages.
- https://raw.githubusercontent.com/langchain-ai/langmem/main/README.md recorded 2026-11-20
LangMem "offers both functional primitives you can use with any storage system and native integration with LangGraph's storage layer"; examples use InMemoryStore and AsyncPostgresStore.
- https://www.langchain.com/pricing recorded 2026-11-20
The pricing page sells LangSmith Platform products such as Observability, Evaluation, Agent Server and Fleet, and lists deepagents, langgraph and langchain under "Open Source Agent Frameworks"; it does not mention LangMem.
Adoption
3 medium confidenceAdoption is counted as monthly PyPI downloads of langmem, the library's only package. A download is an install, often an automated one, rather than an agent running in production.
- https://pypi.org/pypi/langmem/json recorded 2026-11-20
PyPI's langmem 0.0.30 is "Prebuilt utilities for memory management and retrieval.", with the repository README as its description.
- https://pypistats.org/api/packages/langmem/recent recorded 2026-11-20
841,033 downloads of langmem in the last month.
Capability
4 medium confidenceLangMem extracts memories from conversations, consolidates or removes outdated ones with an LLM, and keeps semantic, episodic and procedural memory in namespaces an agent can search per user. It has no time-based forgetting of its own and leaves expiry to the store beneath it.
- https://langchain-ai.github.io/langmem/concepts/conceptual_guide/ recorded 2026-11-20
Memory types are semantic (collections and profiles), episodic and procedural; memory managers "update or remove outdated memories, and consolidate"; memories sit in multi-level namespaces such as per user. The guide names no expiry or decay setting.
- https://raw.githubusercontent.com/langchain-ai/langmem/main/README.md recorded 2026-11-20
Memory management tools let agents "record and search information during active conversations", and a "Background memory manager" "automatically extracts, consolidates, and updates agent knowledge".
Verified 2026-11-20, except some scores, still awaiting re-confirmation