Letta (MemGPT)
Agents with persistent memory — recall past conversations
📋 Overview
MemGPT is an AI agent framework that provides persistent memory — enabling agents to recall past conversations, edit their context, and maintain long-term relationships. Unlike AutoGPT (which has no memory), MemGPT agents remember everything and can reference past interactions. It provides a comprehensive solution for modern AI workflows.
✨ Key Features
- •14K+ GitHub stars — persistent memory agents
- •Recall past conversations
- •Edit agent context
- •Long-term relationships
- •Any LLM support
- •Apache 2.0 licensed
🎯 The Problem It Solves
AI agents forget everything between sessions. MemGPT provides persistent memory that enables long-term relationships and context.
🔧 How It Works
MemGPT uses a memory management system that stores conversations in a vector database. Agents can recall past conversations, edit their context, and maintain long-term relationships. The framework supports any LLM and provides a simple API.
🚀 Installation & Quick Start
Installation
pip install -U lettaQuick Start
- pip install letta
- letta server
- Open localhost:8283
- Create agent and chat
✅ Pros
- •Best persistent memory
- •Recall past conversations
- •Any LLM support
- •Apache 2.0 licensed
- •Active community
- •Comprehensive documentation
❌ Cons
- •Smaller community than AutoGPT
- •Newer tool
- •Documentation sparse
💬 Practitioner Verdict
“MemGPT is the best framework for agents with persistent memory — the memory management is genuinely differentiated. The trade-off: smaller community than AutoGPT, newer tool, and the documentation is sparse. For agents that need long-term memory, MemGPT is the default.”
Self-Hosted (Free)
Open source, MIT/Apache licensed. Run it yourself.
⭐ Star & Clone on GitHubFree forever. Your infrastructure, your data.
📊 Specifications
- Language
- Python
- License
- Apache-2.0
- Platform
- Linux, macOS, Windows
- Supported Models
- REST API, CLI
💰 Pricing Reality
100% free Apache 2.0. No paid tier.