Letta is an AI research lab based in San Francisco focused on building stateful AI agents with persistent memory. The company's work centres on enabling agents to continuously learn and improve from their own experience, with core technical domains spanning AI agents, persistent memory, continual learning, and memory management.
The company's main product, Letta Code, is a model-agnostic agent harness that provides persistent memory, computer use capabilities, skills, subagents, and deployment paths for language models. It is designed to work across different underlying models rather than being tied to a single provider.
Letta places emphasis on transparency and user ownership, allowing developers to see exactly how their agents work at the token level. This approach reflects a broader commitment to giving developers visibility into and control over the behaviour of AI systems they deploy.
