Letta, formerly MemGPT, is a framework for stateful agents that manage their own memory. OwlMem is shared memory for the agents you already use.
Facts checked against each vendor's public docs and pricing pages on October 25, 2026. Features and prices change; if we got something wrong, email [email protected] and we'll fix it.
At a glance
OwlMem vs Letta
Feature
OwlMem
Letta
What it is
Hosted service for teams: knowledge base, memory and vault
Framework for stateful agents (formerly MemGPT)
MCP
Yes: Hosted MCP server with OAuth sign-in
Partly: Hosted MCP to create and message Letta agents
Shared across AI apps
Yes: Claude, ChatGPT, Cursor, Codex, Gemini, any MCP client
Not published: Not documented for other AI apps
Team sharing
Yes: Organizations, members, one identity per AI
Yes: Teams plan; shared memory repositories (cloud)
How data is stored
Markdown in git, plus facts recalled by meaning
Git-backed memory files plus a vector archive
Audit and rollback
Yes: Every write is a git commit: diff, revert, export
Yes: Every memory change is a git commit
Self-host / open source
No: No, managed only (servers in Canada); CLI is MIT
Agents or memory. Letta is where you build and run agents, and memory belongs to those agents. OwlMem is only the shared layer: Claude, ChatGPT, Cursor, Codex and others keep working as they are and connect to it over MCP.
Both use git. Letta's memory filesystem commits every memory change to git, and OwlMem commits every knowledge-base write to git. In both, history is readable and reversible.
MCP. Letta's hosted MCP server lets other clients create and message Letta agents. OwlMem's MCP server gives any client direct tools to search, write and recall shared memory.
Self-hosting. Letta is open source (Apache-2.0) and runs locally. OwlMem is managed only.
Which one to choose
Letta
Choose Letta if you are building long-running, stateful agents that manage and rewrite their own memory, and you want an open-source framework you can self-host.
Choose OwlMem if you want the agents you already use to share one memory and knowledge base, instead of building new agents on a framework.
Using both, or moving over
If your Letta agents can use MCP tools, they can connect to OwlMem like any other client, so your Letta agents and your everyday AI apps read the same knowledge base. Credentials can live in the OwlMem vault instead of in each agent's settings.
Questions
Is Letta the same as MemGPT?
Yes. Letta grew out of the MemGPT research project. It now focuses on Letta Code, an agent harness with persistent memory, and the Letta API.
Do I need to build agents to use OwlMem?
No. OwlMem works with the AI apps you already have — Claude, ChatGPT, Claude Code, Codex, Cursor, Gemini CLI and any MCP client. Tell your AI to read https://mcp.owlmem.com and it sets itself up.
How do the prices compare?
As of October 2026, Letta has a free tier and paid plans from $20 a month, with Teams Pro at $20 per seat a month. OwlMem is free for one AI agent and from $19 a month, counted per AI agent rather than per person. See pricing.