letta
Letta is the platform for building stateful agents: AI with advanced memory that can learn and self-improve over time.
Overview
Letta (formerly MemGPT) is a platform for building stateful AI agents with advanced memory capabilities that can learn and self-improve over time. Unlike traditional chatbots that reset after each conversation, Letta agents maintain persistent memory across interactions, allowing them to build context, learn from past experiences, and evolve their responses. The platform offers two main products: Letta Code, a CLI tool for running agents locally on your computer, and Letta API for integrating stateful agents into applications. Letta supports skills and subagents, enabling complex workflows and specialized capabilities. The platform is model-agnostic, working with various LLM providers, though it recommends Opus 4.5 and GPT-5.2 for optimal performance. With over 21,000 GitHub stars, Letta provides both TypeScript/Node.js and Python SDKs, making it accessible for developers building applications that require persistent AI memory and learning capabilities.
Pros
- + Advanced persistent memory system that allows agents to learn and improve over time across sessions
- + Dual deployment options with both local CLI tool and cloud API for different use cases and security requirements
- + Model-agnostic architecture supporting multiple LLM providers with extensive SDK support for TypeScript and Python
Cons
- - Requires Node.js 18+ for CLI usage, which may limit adoption in some environments
- - API-based functionality requires API keys and cloud dependency for full feature access
- - As a relatively new platform for stateful agents, may have a learning curve for developers new to persistent memory concepts
Use Cases
- • Building coding assistants that remember project context and learn from previous debugging sessions
- • Creating customer support agents that maintain conversation history and learn customer preferences over time
- • Developing personal AI assistants that evolve their responses based on user behavior patterns and feedback