Skip to content
Memorg

← All comparisons · Compare

Memorg vs Letta (formerly MemGPT)

Stateful agent runtime · OS-inspired memory

Letta and Memorg solve adjacent problems. Letta is an agent runtime that includes a memory abstraction. Memorg is a memory library that does not run agents.

Feature Memorg Letta (formerly MemGPT) Advantage
Scope Memory library Agent runtime with memory built in Comparable
Memory model Hierarchical (session/conversation/topic/exchange) + memory items Core memory + recall memory + archival memory Comparable
Who decides what to recall Deterministic scorer over semantic + recency + importance The agent itself, via tool calls Memorg
Runs an agent loop No Yes Letta (formerly MemGPT)
Deployment In-process Python library Letta server (long-running) Memorg
Storage SQLite + USearch (one file) Postgres + vector store (server) Memorg
MCP integration First-party MCP server Available via runtime Comparable
Best for Apps adding memory to an existing stack Greenfield agents that want a runtime Comparable
License MIT Apache-2.0 Comparable

Pick Memorg when

  • You already have an agent loop, server, and orchestrator you like
  • You want a memory layer you can drop into a non-agent app (RAG, chat, search)
  • You want deterministic retrieval ranking, not an LLM deciding what to read
  • You want a single SQLite file rather than a long-running stateful service

Pick Letta (formerly MemGPT) when

  • You want the runtime, the memory model, and the deployment story in one package
  • You like the OS-inspired core / recall / archival memory model and self-editing memory
  • You want first-class long-running agents with server-managed state
  • You want web-based agent debugging and inspection tooling out of the box

Still deciding?

Most teams end up running both kinds of system in different parts of their stack. Try Memorg where the answer is "I just need recall over a session," and keep Letta (formerly MemGPT) where you need the rest of its surface.