MemGPT is now part of Letta
MemGPT started as a research project at UC Berkeley, inspired by an effort to improve the memory systems of Discord chatbots. As it turns out, many builders in the AI ecosystem were (and still are!) looking for a memory management solution, and the paper we wrote went viral. Since then, the scope of our research and open source development has grown significantly, and the MemGPT repository has rapidly evolved to support agents more generally, with support for features like custom tools, data sources, and even memory classes, as well as the ability to deploy and persist these agents as stateful services.
The rapid popularity of the research, the paper, and the context window management technique itself has resulted in some confusion about the MemGPT name. Some people use “MemGPT” to refer to the original “LLM OS” techniques in the paper, some use “MemGPT” to refer to an archetype of LLM-driven chatbots with self-editing memory, and others use “MemGPT” to refer to the open source agent framework. To clarify this, we’ve decided that MemGPT should refer to the original agent design pattern described in the research paper (empowering LLMs with self-editing memory tools), and use the name Letta to refer to the agent framework.
Introducing Letta, the company we’ve started to advance the frontier of AI systems with memory. We started Letta to continue research and development of the LLM OS: the orchestration layer that sits above the model layer. Letta the company will continue to maintain the open source repository, MemGPT. Our commercial offering will focus on making deployability easier, and improving performance, and offering UX tools for debugging and monitoring agents.
This means:
- The pypi package will be moved to
letta
- The Docker image will be moved to
letta/letta-server
Thank you to our open source and Discord communities for using MemGPT and invaluable feedback over the course of the project. We hope to keep having you along for this ride, and promise to fix bugs faster now that we are not just two PhD students :)
- Charles and Sarah
Traditional LLMs operate in a stateless paradigm—each interaction exists in isolation, with no knowledge carried forward from previous conversations. Agent memory solves this problem.
As AI agents become more sophisticated, understanding how to design and manage their context windows (via context engineering) has become crucial for developers.
Today we're announcing Letta Filesystem, which provides an interface for agents to organize and reference content from documents like PDFs, transcripts, documentation, and more.
Sleep-time compute is a new way to scale AI capabilities: letting models "think" during downtime. Instead of sitting idle between tasks, AI agents can now use their "sleep" time to process information and form new connections by rewriting their memory state.
We've releasing new client SDKs (support for TypeScript and Python) and upgraded developer documentation
Introducing Agent File (.af): An open file format for serializing stateful agents with persistent memory and behavior.
Introducing “stateful agents”: AI systems that maintain persistent memory and actually learn during deployment, not just during training.
Introducing the Letta Agent Development Environment (ADE): Agents as Context + Tools
