Context Constitution
Today we're releasing the Context Constitution: a set of principles governing how AI agents manage context to learn from experience.
Letta Code — Memory-first agents that learn
Create agents that can be taught with natural language and improve from experience
Persistent agents instead of stateless sessions
Build your own deeply personalized agents, each with their own unique experiences and persona, designed to evolve with you and your work.
Always improving and learning
Letta Code uses background memory subagents to improve prompts, context, and skills over time with experience. View your agent's memory in the memory palace.
Own your memory and port it across models
Transparently view and modify your agent's memory and context. Port memory and conversation histories between models, even from different providers.
Chat from any device, run on any environment
Use letta server to remote control agents running on external devices. Move agents across machines along with their memory, context, and files.
AI Research
Building machines that learn
Letta is an AI lab building machines the learn: persistent agents with the ability to continuously learn and adapt from their own experience. Born from MemGPT at UC Berkeley's Sky Computing Lab, and backed by leaders in AI research and infrastructure.
Research blog }
Introducing Context Repositories: Git-based Memory for Coding Agents
We're introducing Context Repositories, a rebuild of how agent memory works, using programmatic context management and git-based versioning.
Continual Learning in Token Space
At Letta, we believe that learning in token space is the key to building AI agents that truly improve over time.
Skill Learning: Bringing Continual Learning to CLI Agents
Today we're releasing Skill Learning, a way to dynamically learn skills through experience.
Context-Bench: Benchmarking LLMs on Agentic Context Engineering
Context-Bench evaluates how well language models chain file operations, trace relationships, and manage long-horizon information retrieval.
Sleep-time Compute: AI that Dreams
We introduce sleep-time compute, a technique that allows agents to reason about context during idle time rather than at inference.
Your personalized agent, accessible from anywhere
Letta Code
The memory-first agent. Try for free with your own API keys or existing coding plans.
