Letta is a memory management system for AI agents, and this server exposes it through MCP so Claude can maintain persistent memory across conversations. You'd reach for this when you need your AI interactions to remember context, user preferences, or ongoing work beyond a single session. The server connects to Letta's infrastructure to store and retrieve structured memory, letting you build conversational experiences that actually remember what happened yesterday or last week. If you're building any kind of agent that needs to recall past interactions or maintain state over time, this gives you a standardized way to handle that without rolling your own memory layer.
claude mcp add --transport stdio letta-ai-memory-mcp -- npx -y @letta-ai/memory-mcp