Letta is an AI research lab in San Francisco that builds machines that learn. It develops experiential agents designed to remember everything, learn continuously, and improve themselves over time.
The product ships as Letta Agent, a self-improving AI agent whose memory, identity, and capabilities evolve with experience. Its research explores memory models trained with memory-native reinforcement learning, principles for how agents manage context to learn from experience, git-based memory with programmatic context management and versioning, continual learning in token space, and sleep-time compute that allows agents to reason about context during idle time rather than at inference. The original foundation traces to MemGPT, an approach to virtual context management treated as an LLM operating system.
Installation uses the command npm i -g @letta-ai/letta-code. The work has been featured in case studies with Bilt, 11x, Kognitos, and Hunt Club. Coverage in outlets such as WIRED, Fast Company, and a video by Matthew Berman has discussed its concepts including sleep-time compute.
Letta sits in PulseGate's AI & ML category. It focuses on building AI agents that maintain long-term memory, learn from experience, and evolve their capabilities over time. Letta is an open-source project aimed at developers and AI researchers. Letta is open source under the Open Source license. It runs on the web and the command line, and it can be self-hosted.
Letta AI builds and maintains Letta, and it first shipped in 2023. Among its 5 catalogued features are Stateful Agents, Persistent Memory, and Continual Learning. It exposes integrations via a public API.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Same category — not a similarity match