Letta (formerly MemGPT) for Long-Lived Agents
Letta specializes in agents that live for days or weeks with self-managed memory.
Letta (formerly MemGPT) for Long-Lived Agents is a free AI Agents lesson on CoddyKit — lesson 2 of 4. You can read the complete lesson below for free — then practise it hands-on in the browser with a built-in code editor and a 24/7 AI tutor. It is part of the AI Agents learning path, one of 4 lessons in the course, and your progress syncs across the web and the CoddyKit app.
What Is Letta?
Letta (formerly MemGPT, Berkeley 2023) is a framework specifically for agents that live across many sessions and accumulate memory over weeks or months.
The Memory Problem
Standard LLM context windows are 100k-2M tokens. After many sessions, even huge contexts overflow. Letta solves this with explicit memory management.
Two-Tier Memory
- Core memory — always in context (persona, key facts)
- Archival memory — stored in vector DB, retrieved on demand
Self-Managed Memory
The agent itself decides what to remember and what to retrieve. Tools include:
core_memory_append— add to core memoryarchival_memory_insert— save to archivearchival_memory_search— retrieve from archive
Why Self-Managed?
Humans don't remember every detail of every conversation. They distill what matters. Letta's agents do the same — actively curating their memory across sessions.
Setup
# pip install letta
from letta import create_client
client = create_client(base_url='http://localhost:8283')
agent = client.create_agent(
name='Alice',
persona='I am Alice, a helpful assistant.',
human='User\'s name is Bob.',
model='gpt-4o-mini'
)Sending Messages
response = client.send_message(
agent_id=agent.id,
role='user',
message='Remember: I love hiking.'
)
print(response.messages)
# Agent may invoke core_memory_append to save this factInspecting Memory
memory = client.get_in_context_memory(agent_id=agent.id)
print(memory.persona, memory.human)
archival = client.get_archival_memory(agent_id=agent.id)
for m in archival:
print(m.text)Long-Lived Agents in Production
For chat assistants used over weeks (think a personal coach):
- Standard frameworks: lose context between sessions
- Letta: persistent memory, can recall last week's conversation
Self-Editing Persona
Letta agents can edit their own persona — e.g. update their understanding of who they are over time. Interesting research direction.
Self-Hosting
# Letta runs as a server you can self-host
docker run -d -p 8283:8283 -e LETTA_PG_URI='postgresql://...' letta/letta:latestWhen to Use Letta
- Personal assistants
- Customer-support agents that remember user history
- Tutors / coaches
- Any agent expected to retain knowledge across sessions
When Not To
- One-off tasks (research bots, code-gen)
- Multi-tenant where context isolation is hard
- When simple KV memory + RAG suffices
Comparison to Mem0
Mem0 is a competitor in the long-term memory space. Lighter, less opinionated. Good fit if you want to bolt memory onto existing agents.
Letta Specialty
What is Letta specifically built for?
Recap
Letta = memory-first agent framework. Two-tier memory (core + archival), self-managed by the agent itself. Pick for long-lived assistants; skip for one-off tasks.
Frequently asked questions
Is the “Letta (formerly MemGPT) for Long-Lived Agents” lesson free?
Yes — the full text of “Letta (formerly MemGPT) for Long-Lived Agents” is free to read here on the web, and the AI Agents course includes 4 lessons in total. To practise it interactively (a built-in code editor and a 24/7 AI tutor) and unlock the rest of the AI Agents course, upgrade to CoddyKit PRO.
What will I learn in “Letta (formerly MemGPT) for Long-Lived Agents”?
Letta specializes in agents that live for days or weeks with self-managed memory. You practise AI Agents with hands-on code you run directly in the browser, and a 24/7 AI tutor answers your questions as you work through the lesson.
Do I need any experience to start AI Agents?
No prior experience is required. AI Agents on CoddyKit is structured for beginners through advanced learners; this is — lesson 2 of 4, so you can start here or from the beginning and move at your own pace.
How long does the “Letta (formerly MemGPT) for Long-Lived Agents” lesson take?
Most CoddyKit lessons take about 5–10 minutes. Each one is bite-sized and interactive, so you make steady progress and pick up exactly where you left off across the web and the app.
Can I write and run code in this AI Agents lesson?
Yes. Every AI Agents lesson includes a built-in code editor, so you write and run real code right in your browser and get instant AI feedback — no local setup required.
All lessons in this course
- LangGraph vs CrewAI vs AutoGen
- Letta (formerly MemGPT) for Long-Lived Agents
- OpenAI Assistants API and Threads
- Choosing the Right Framework Per Use Case