Letta is an AI research lab in San Francisco building machines that learn. We envision a future where humans coexist with digital people: experiential agents that remember everything, learn continuously, and improve themselves over time.
Powering agent learning through memory models trained with memory-native RL.
A set of principles governing how AI agents manage context to learn from experience.
A rebuild of how agent memory works, using programmatic context management and git-based versioning.
Learning in token space is the key to building AI agents that truly improve over time.
Letting agents reason about context during idle time, rather than at inference.
The original work on virtual context management — where Letta began.
All research →$ npm i -g @letta-ai/letta-code
Our research ships as software: Letta Agent, a self-improving AI agent whose memory, identity, and capabilities evolve with experience.
Do large language models dream of AI agents?
Why sleep-time compute is the next big leap in AI
Sleep-time compute — AI that “thinks” 24/7
Letta, one of UC Berkeley’s most anticipated AI startups, has just come out of stealth
Founded by the creators of MemGPT from UC Berkeley’s Sky Computing Lab, the birthplace of Spark and Ray. Advised by Ion Stoica and Joey Gonzalez. Backed by Jeff Dean, Clem Delangue, Cris Valenzuela, Jordan Tigani, Tristan Handy, Barry McCardel, Robert Nishihara, Sunflower Capital, Essence VC, and Felicis Ventures.
We’re a small team of researchers and engineers building machines that learn — join our team.
