What is Persistent memory?

Persistent memory is the property of agent memory that distinguishes it from the model's transient context window: the data lives in durable storage (a database, a vector store, a memory server) and remains available after the current session closes. A model's context window is wiped the moment a conversation ends, so anything the agent learned, the shape of a codebase, a user's preferences, a decision and its rationale, vanishes unless it was written somewhere lasting. Persistent memory captures those facts as they arise and retrieves the relevant ones back into context on later tasks, so the agent improves over time rather than relearning the same things every session. The retrieval step matters as much as the storage: because durable stores grow large and the context window stays small, a good memory system selects only what is relevant to the current task rather than dumping everything back in. Delivered as an MCP server, persistent memory becomes a tool the model calls to record and recall context. Glen implements exactly this, durable, organization-scoped memory retrieved on demand through a single MCP tool.