letta-ai/letta

Platform for stateful agents: AI with advanced memory that can learn and self-improve over time.

What it solves

Letta 提供了一种方式来构建具备高级记忆的 AI 代理，这些代理能够随时间学习并自我改进，突破标准短期上下文窗口的限制。

How it works

Letta 允许开发者创建保持持久记忆块（例如人格和人类细节）的有状态代理。它与模型无关，可通过 CLI 工具用于本地任务，或通过 Python 与 TypeScript SDK 的 API 进行应用集成。系统支持使用技能和子代理来增强代理能力。

Who it’s for

想要在应用中集成有状态、自我改进的 AI 代理，或在终端本地运行自主代理以协助编码和一般计算任务的开发者。

Highlights

  • Advanced Memory: Enables agents to learn and maintain state over time.
  • Model Agnostic: Works with various LLMs, including recommendations for high-performance models like Opus 4.5 and GPT-5.2.
  • Flexible Deployment: Offers both a local CLI for terminal-based agents and a full-featured API for app development.
  • Extensible: Supports the addition of skills and subagents for complex workflows.