Agent File
Today we are launching Agent File (.af): a file format to make stateful agents shareable and reproducible. With Agent File, you can re-create the exact same agent on a different server – allowing you to move agents across cloud and local agent servers.
While sharing prompts is fairly straightforward, stateful/autonomous agents have significant associated state: tools definitions and settings, the execution environment, model configuration, and memories (both in-context and external).
Agent File was build to address key challenges with the agent ecosystem today:
- Portability and reproducibility between systems
- Collaboration between developers
- Preservation of agent configurations
- Versioning agents over time
With the release of Agent File, we’re also sharing a few example agents we built, including:
- A Deep Research inspired agent that uses in-context memory to manage research state over time
- A customer support agent with memory and custom tools
- A workflow agent (stateless with a pre-defined tool graph) for drafting emails
- MemGPT agents with prompts from the original release
You can download an .af to instantly create each of these agents (we’ve also released scripts to generate the agents as well).
A key part of building towards stateful agents is determining what the standard for “state” in an agentic system should be: Agent File is the first file format for representing agentic state in a portable and reproducible way. This is a critical step toward our vision of building the operating system for AI agents – systems that can truly learn and improve with experience.
We're excited to see what you build! Try it out and let us know what agent types you'd like to see next, or share your own agent creations with the community.
Check out our example agent files and upload your own on the Agent File GitHub repository.
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.
Although RAG provides a way to connect LLMs and agents to more data than what can fit into context, traditional RAG is insufficient for building agent memory.
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.
We've releasing new client SDKs (support for TypeScript and Python) and upgraded developer documentation
Introducing the Letta Agent Development Environment (ADE): Agents as Context + Tools
Letta v0.6.4 adds Python 3.13 support and an official TypeScript SDK.
Letta v0.5.2 adds tool rules, which allows you to constrain the behavior of your Letta agents similar to graphs.
Letta v0.5.1 adds support for auto-loading entire external tool libraries into your Letta server.
We're excited to announce the Letta Leaderboard, a comprehensive benchmark suite that evaluates how effectively LLMs manage agentic memory.
Memory blocks offer an elegant abstraction for context window management. By structuring the context into discrete, functional units, we can give LLM agents more consistent, usable memory.
