What is Agent memory?

Agent memory is the layer that lets an AI agent carry knowledge across conversations rather than forgetting everything when a session ends. Without it, each new chat starts from zero: the agent re-learns the codebase, re-asks the same questions, and repeats mistakes a teammate's agent already made. A memory system captures durable facts, decisions, and preferences as they come up, then surfaces the relevant ones back into context on later tasks. Memory can be scoped per user, but the higher-leverage design is org-shared memory, where every agent on a team reads and writes the same store, so a fact one engineer's agent learns is instantly available to everyone else's. Delivered as an MCP server, agent memory becomes a tool the model calls to recall and record context in the same round trip, which is how Glen approaches it: shared memory for AI agents, scoped to the organization, available through one MCP tool.