What is Letta?
Letta, formerly known as MemGPT, supports AI agents that retain identity and long-term memory across conversations, models, and computers. Developers can embed agents in TypeScript applications through the Letta Agent SDK or interact with them through a terminal interface and a browser. Supported messaging channels include Slack, Telegram, Discord, and custom channels.
The actively developed runtime lives in the Letta Code repository and includes the agent harness, terminal interface, App Server, and channels. The App Server supports local or self-hosted agents, while Letta Cloud keeps agent memory, identity, and conversations available across computers. The original Letta repository retains historical source for a retired API server rather than the current implementation.
Letta describes itself as an AI research lab in San Francisco. Letta Code is its stateful agent harness for agents that have memory, identity and a sense of experience over time. Agents can be used through a local CLI, through the browser at chat.letta.com including on mobile, and through messaging integrations.
Key capabilities
- Persistent memory Agents retain identity and long-term memory across conversations, models, and computers.
- Terminal interface Letta Code provides an interactive terminal interface for working with agents.
- Local deployment The App Server runs agents locally or in a self-hosted environment.
- Application development The Letta Agent SDK embeds stateful agents in TypeScript applications.
- Messaging channels Agents can be accessed through Slack, Telegram, Discord, and custom channels.
- Cloud continuity Letta Cloud keeps agent memory, identity, and conversations available across computers.
Where Letta fits
Letta fits applications and conversational workflows that need an agent to retain context beyond an individual session. Its SDK supports embedding agents in TypeScript applications, while terminal, browser, and messaging access provide different ways to interact with them.
Before deployment, teams should distinguish the current Letta Code runtime from the retired API server retained in the original repository. They should also verify deployment requirements, memory retention controls, channel permissions, hosting location, and cloud costs for their intended configuration.
Install
npm install -g @letta-ai/letta-codeReviews of Letta
Be the first to share your experience with Letta.
Write a review
Loading review form…
Alternatives to Letta
LangChainLangChain is an MIT-licensed open-source framework (Python/JS) for building LLM applications and agents via LangGraph; LangSmith adds optional hosted tracing, evaluation and deployment.
CrewAICrewAI is an MIT-licensed open-source Python framework for orchestrating multi-agent AI workflows; CrewAI Enterprise adds a hosted runtime with a free Basic tier and custom-priced governance features.