Letta (MemGPT)
Letta (formerly MemGPT) is a framework for building stateful AI agents with memory that can learn and improve over time. It provides an agent harness, interactive terminal UI, App Server, and channels for deployment. It is for developers who want to create agents with persistent memory across sessions.
✨ Key features
- Stateful agents with memory that can learn and improve
- Interactive terminal UI for local agent interaction
- App Server for local or self-hosted agents
- Deploy to Slack, Telegram, Discord, and custom channels
- Agent SDK for building agents into TypeScript applications
- Letta Cloud for cross-device memory and identity
🎯 Use cases
- Build a personal assistant that remembers user preferences over time
- Create a customer support bot that retains conversation history
- Develop a coding agent that learns from past interactions
- Deploy agents to messaging platforms like Slack or Discord
- Integrate stateful agents into TypeScript applications
📦 Installation
🧰 Requirements: Requires Node.js and npm to install the package; no API keys or accounts mentioned for basic usage.
npm install -g @letta-ai/letta-code
🚀 Usage
Launch the interactive terminal UI:
letta
Run the App Server:
letta server
⚠️ Good to know
The current repository is a landing page; the active source is in letta-ai/letta-code, and the archived V1 server is unsupported and not for production use.
❓ FAQ
How do I install Letta?
Install globally via npm with npm install -g @letta-ai/letta-code.
How do I start the interactive terminal UI?
Run letta in your terminal after installation.
Can I run Letta as a server?
Yes, run letta server to start the App Server for local or self-hosted agents.
What happened to the old MemGPT repository?
The old V1 server source is archived on the archive branch and is unsupported; active development is in letta-ai/letta-code.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.