Letta is an open-source platform for building stateful AI agents with persistent memory. The project grew from research around MemGPT and focuses on agents that can maintain information, context, and identity across long-running interactions instead of treating every model call as an isolated conversation.

Letta provides APIs and developer tooling for defining agents, connecting models and tools, managing memory, and running agent services. Its architecture makes memory an explicit part of the agent system, enabling applications to inspect, update, and persist state rather than depending only on a model's temporary context window.

The platform is aimed at developers building assistants and autonomous systems that need continuity over time. It is particularly relevant for personal agents, research agents, and applications where durable knowledge and evolving state are central to the user experience.