Letta | Machines that learn

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.

Viva la machina.

RESEARCH

Memory Models: Towards Agents That Learn JUN 2026
Powering agent learning through memory models trained with memory-native RL.

Context Constitution APR 2026
A set of principles governing how AI agents manage context to learn from experience.

Context Repositories: Git-based Memory FEB 2026
A rebuild of how agent memory works, using programmatic context management and git-based versioning.

Continual Learning in Token Space DEC 2025
Learning in token space is the key to building AI agents that truly improve over time.

Sleep-time Compute: Learning Offline APR 2025
Letting agents reason about context during idle time, rather than at inference.

MemGPT: The LLM Operating System OCT 2023
The original work on virtual context management — where Letta began.

PRODUCT

$ 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.

Featured case studies: Bilt, 11x, Kognitos, Hunt Club.

PRESS

• WIRED
  Do large language models dream of AI agents?
• Fast Company
  Why sleep-time compute is the next big leap in AI
• Matthew Berman
  Sleep-time compute — AI that “thinks” 24/7
• TechCrunch
  Letta, one of UC Berkeley’s most anticipated AI startups, has just come out of stealth

TEAM

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.