How is this different from LangChain's ConversationBufferMemory and vector stores?

Those are per-app memory you wire and maintain yourself. Glen is durable, org-shared long-term memory delivered as one MCP tool, read and written by every agent across your organization and frameworks.

Do I still need an embedding pipeline?

No. Glen handles relevance and storage behind one MCP tool, so you skip choosing a vector store, running embeddings, and building separate read and write paths.

Shared memory for LangChain

LangChain gives you the building blocks for LLM applications, chains, tools, retrievers, and agents that compose models with the outside world. For memory it ships conversation buffers and a long list of vector-store integrations, but that leaves you owning the hard part: choosing a store, running an embedding pipeline, and deciding how durable knowledge gets written, deduplicated, and shared across chains and deployments. Glen, shared memory for AI agents, gives your LangChain agents long-term shared memory as a single MCP tool, so every chain in your organization reads from and writes to one store instead of each one carrying its own bolted-on memory.

Wire Glen into LangChain over MCP and any agent or chain can call one tool that both retrieves relevant long-term context and records new facts in a single round trip. Instead of stitching together a vector store, an embedding model, and a retrieval chain just to give an agent durable recall, you point it at Glen and let the memory layer handle relevance and writes for you. A retrieval step pulls what your organization already knows before the model acts; a later step writes back what the run learned.

The difference from LangChain's built-in memory is scope and persistence. ConversationBufferMemory and the vector-store retrievers you assemble are per-app and per-process; you wire each one and keep them in sync yourself. Glen is org-scoped, so the same memory is shared across every chain, every agent, and every deployment in your organization. What one LangChain agent learns, every other agent learns, including agents built on entirely different frameworks. And because Glen is a standard MCP server, that shared memory is also readable from Claude Code, Cursor, or any other MCP client, so the knowledge your chains accumulate is the same knowledge your developers and other tools draw on. You connect once over OAuth or an API key and the memory compounds with every run.

FAQ

How is this different from LangChain's ConversationBufferMemory and vector stores?: Those are per-app memory you wire and maintain yourself. Glen is durable, org-shared long-term memory delivered as one MCP tool, read and written by every agent across your organization and frameworks.
Do I still need an embedding pipeline?: No. Glen handles relevance and storage behind one MCP tool, so you skip choosing a vector store, running embeddings, and building separate read and write paths.