How is this different from AutoGen's teachability and memory?

AutoGen's teachable agents and memory stores help a single agent recall across its own sessions. Glen is org-wide shared memory that every agent, every group chat, and every other MCP client in your organization reads and writes through one tool.

How do my AutoGen agents reach Glen?

Glen is a standard MCP server. Register it as a tool on any agent, then call it during a conversation to retrieve context and record observations in one round trip.

Shared memory for AutoGen

AutoGen is built around conversable agents that talk to each other, an assistant, a user proxy, a group chat manager passing messages until the task is done. That conversation is where the reasoning happens, but it is also where the knowledge dies: once the chat terminates, the message history is gone unless you serialize it somewhere, and the next conversation starts with an empty context window. AutoGen's teachability and memory add-ons help a single agent remember within its own lifecycle, but cross-conversation, cross-agent knowledge for your whole organization is still something you stitch together yourself. Glen, shared memory for AI agents, gives your AutoGen agents that durable org-wide memory as a single MCP tool every agent reads from and writes to.

Register Glen as a tool on your AutoGen agents and any participant in a conversation gains one capability that both retrieves relevant long-term context and records new facts in a single round trip. Before a group chat kicks off, an agent can pull what the organization already knows about the task or entity so the conversation starts informed instead of cold; as the chat produces durable findings, an agent writes them back without you building a serialization layer or a vector store to hold them.

AutoGen's model is many agents conversing, and Glen extends that conversation across time and across separate runs. Because the memory is org-scoped rather than pinned to one agent's teachable memory or one chat's message list, a fact discovered in today's group chat is available to a different team of agents tomorrow, even one assembled for a different task or running in a different framework. The knowledge compounds instead of resetting at every `initiate_chat`. And because Glen is a standard MCP server, the same store your AutoGen agents write is readable from Claude Code, Cursor, or any other MCP client, so automated multi-agent runs and human-driven work share one source of truth. Connect once over OAuth or an API key and let the memory grow as a byproduct of every conversation.

FAQ

How is this different from AutoGen's teachability and memory?: AutoGen's teachable agents and memory stores help a single agent recall across its own sessions. Glen is org-wide shared memory that every agent, every group chat, and every other MCP client in your organization reads and writes through one tool.
How do my AutoGen agents reach Glen?: Glen is a standard MCP server. Register it as a tool on any agent, then call it during a conversation to retrieve context and record observations in one round trip.