Letta, formerly MemGPT, came out of research into giving agents memory larger than the context window. The agent decides what to keep, what to page out and what to recall, so a conversation can run for months.
Reach for it when
- Your agent must remember a user across months
- Context window limits are the thing breaking you
- You want memory as a first class feature, not a bolt on
Look elsewhere when
- A single conversation is all you need
- You want the simplest possible setup
What the lessons will cover
- 01Install and setup
- 02Agents and memory blocks
- 03Core and archival memory
- 04Tools
- 05The server and REST API
- 06Multi agent
- 07Persistence
- 08A long lived assistant
Others in autonomous
AutoGPTGive it a goal and walk away.
AgnoFast, lightweight agents with memory built in.
GriptapeStructured agents, pipelines and workflows.
smolagentsA tiny agent library that writes Python.Common questions
Is Letta free to use?
Letta is open source and free to run yourself. You still pay whichever model provider you point it at, and this tutorial is free with no signup.
Do I need to know Python to use Letta?
Basic Python is enough. Functions, dictionaries and imports cover most of what Letta asks of you.
When should I not use Letta?
A single conversation is all you need. You want the simplest possible setup.
Lessons for Letta are being written. Meanwhile the LangGraph tutorial covers the same ground: state, tools, loops, memory and human approval. Most of it carries straight over.
Start the LangGraph tutorial →