What is Memory store?
A memory store is the durable backend where an AI agent's long-term memory actually lives, the database or service that persists facts and observations and serves the relevant ones back into context on demand.
A memory store is the concrete place an agent's long-term memory is kept and queried. Where agent memory and persistent memory name the capability, the memory store is the system that implements it: it persists each recorded fact or observation, organizes them so they can be found again, and retrieves the relevant subset when a later task asks for it. Memory stores take several shapes. Some are vector databases that embed each memory and retrieve by semantic similarity; others are structured stores, often relational, that organize memories into named collections and select by relevance. A good store is more than a bucket: it handles scoping (which memories belong to which user, agent, or team), provenance (who or what recorded a fact), and the retrieval logic that keeps the small context window filled with only what matters rather than the whole store. When a store is shared across an organization, every member's agent reads and writes the same memory, so knowledge compounds across the team. Exposed through an MCP server, the memory store becomes a tool the agent calls to record and recall context. Glen organizes memory into org-owned stores, each holding shards and observations, every member can read and write, reached through one MCP tool.