Letta构建拥有长期记忆和连续学习能力的状态AI代理框架

5.0 (6)

审阅者 Daniel Nikulshyn·更新 2026年5月

Developer Tools AI Agents Open Source SDK Long-Term Memory Framework API Multi-LLM

概览

Letta 是一个开发者平台，帮助创建能够跨会话保留上下文、从交互中学习并随时间改进行为的 AI 代理。不同于无状态的聊天机器人，Letta 代理拥有持久记忆，能够回忆起过去的对话、用户偏好以及积累的知识。该框架为管理代理的记忆、推理和工具使用提供基础设施，并支持多家LLM提供商。开发者可以通过SDK和可视化界面构建、部署并监控代理，适用于个人助理、客户支持以及受益于连续性的自主工作流程等应用场景。

主要功能

拥有持久内存的状态代理
自编辑内存块
支持多个 LLM 提供商
工具和函数调用
代理开发环境（ADE）
REST API 和 Python/TypeScript SDK

价格

模型: Free
分类: AI Agent Memory
评分: 5.0 / 5 (6)

使用场景

拥有记忆的个人 AI 助手

创建能够记住用户偏好、过去的对话和会话间上下文的助手，与此同时提供更个性化和持续的交互。

具备上下文意识的客户服务代理

部署客户代理，能够回忆客户的历史、先前票据和累积知识，从而解决问题时不需要用户重复说明。

无人操作工作流自动化

创建具有使用工具调用的状态代理，可在执行多步workflow 时保持状态并从之前运行中学习以改善可靠性。

代理原型化和调试

使用代理开发环境和 SDK，可视化地检查内存块、推理和工具使用，同时在状态代理行为上迭代。

优点 & 缺点

优点

会话之间保持长期的持久内存
模型中立，支持多个 LLM 提供商
开源基础，持续开发
可视化工具用于检查代理状态和内存
开源基础，持续开发
可视化工具用于检查代理状态和内存

缺点

需要技术设置和开发人员的专家知识
内存管理增加了简单的 LLM 调用上的复杂性
与主流代理框架相比，稍小的生态系统

评测

5.0

6 个评分的平均值。

登录以留下评测。

Elena Rossi

May 7, 2026

Use it every day

Honestly didn't expect to like it this much. Stateful agents with persistent memory is exactly what I needed, and visual tools for inspecting agent state and memory. I do wish memory management adds complexity over simple LLM calls, but I reach for it almost every day now and it just clicks.

Esther Adeyemi

Apr 14, 2026

Does the job

Pretty happy overall. Stateful agents with persistent memory just works and open-source foundation with active development. but no dealbreakers — I'd recommend it to a friend without hesitating.

George Papadakis

Dec 4, 2025

Skeptical, then convinced

I went in skeptical — most tools in this space overpromise. It actually delivers on self-editing memory blocks, and visual tools for inspecting agent state and memory caught me off guard. still, I'd recommend giving it a real trial.

Wei Chen

Sep 21, 2025

Years in this space

I've evaluated a lot of these over the years. What stands out here is rEST API and Python/TypeScript SDKs — handled better than most — and persistent long-term memory across sessions. Memory management adds complexity over simple LLM calls is my one real gripe. Worth the time if this is your use case.

Joanna Kowalski

Aug 12, 2025

Skeptical, then convinced

I went in skeptical — most tools in this space overpromise. It actually delivers on tool and function calling, and visual tools for inspecting agent state and memory caught me off guard. Requires technical setup and developer expertise is why this isn't a perfect score, still, I'd recommend giving it a real trial.

Ethan Brooks

Jul 13, 2025

Solid for our team

We rolled this out across the team last quarter and visual tools for inspecting agent state and memory. Self-editing memory blocks fits neatly into how we already work, and tool and function calling removed a step we used to do by hand. but it has held up under daily use.