Does Glen conflict with Letta's own memory system?

No. Letta manages each agent's working and archival memory; Glen adds a separate, org-shared long-term store reachable over MCP. They are complementary, Letta for per-agent state, Glen for cross-agent knowledge.

How do Letta agents use Glen?

Connect Glen as an MCP tool. Any Letta agent can then call it to retrieve org-shared context and write observations that every other agent and MCP client can read.

Shared memory for Letta

Letta (formerly MemGPT) is built around the idea that agents should have memory, with a clever architecture for managing what stays in context and what gets paged to a longer-term store. That memory, though, is fundamentally per-agent: each Letta agent owns its own memory blocks and archival store, and there is no built-in way for a fleet of agents, or your other tools, to share one organizational source of truth. Glen, shared memory for AI agents, complements Letta by adding an org-scoped memory layer over MCP, so many Letta agents, and the rest of your stack, can read from and write to one shared store.

Letta is excellent at managing a single agent's working and archival memory. Glen solves a different problem: the knowledge that should be shared across agents and across frameworks. Expose Glen to your Letta agents as an MCP tool and any agent can call it to retrieve what the organization already knows and to record durable facts that other agents, not just itself, will see. Letta keeps doing what it does well for each agent's own state; Glen becomes the common ground between them.

Because Glen is org-scoped, a fact one Letta agent learns is immediately available to every other Letta agent and to any other MCP client, instead of being trapped in one agent's archival memory. That is the gap a per-agent memory architecture leaves: cross-agent, cross-deployment sharing. A support agent records a durable customer constraint; a separate analysis agent, or a human in Claude Code or Cursor, reads it next time. And because Glen speaks standard MCP, you do not have to bridge memory formats between frameworks, the same store is reachable from anywhere that speaks MCP. Wire it in once over OAuth or an API key and let your Letta fleet share one growing memory instead of many isolated ones.

FAQ

Does Glen conflict with Letta's own memory system?: No. Letta manages each agent's working and archival memory; Glen adds a separate, org-shared long-term store reachable over MCP. They are complementary, Letta for per-agent state, Glen for cross-agent knowledge.
How do Letta agents use Glen?: Connect Glen as an MCP tool. Any Letta agent can then call it to retrieve org-shared context and write observations that every other agent and MCP client can read.