letta
Letta (formerly MemGPT) builds stateful agents with memory that learn and improve over time. Its agents keep memory, identity and a history of experience, and rewrite their own context, including memory blocks and system prompts, as they learn. Note that development has moved: the main letta-ai/letta repository now points to letta-ai/letta-code, which contains the agent harness, terminal UI, App Server and channels. The original Letta V1 API server is retired and kept on an archive branch.
Letta Code is aimed at people who want long-lived agents rather than one-off chat sessions: a personal or team assistant that remembers past work, agents that run on schedules, or agents that coordinate subagents. You can talk to the same agent from the CLI, a desktop app, the browser (chat.letta.com), Slack, Telegram or Discord. Context, including memory, is tracked with git and can be synced to your own repository, and skills, hooks and permission modes control behaviour.
Getting started: install with `npm install -g @letta-ai/letta-code` and run `letta` in a project directory. On first launch choose Letta Cloud or local mode, then use `/connect` to add your own model API keys and `/model` to switch models. `letta server` runs the App Server for local or self-hosted agents.
Features
Why choose it
Trade-offs
Maintenance check
- Last commit
- 2026-10-18
- Commits, last 90 days
- 9
- Releases, last 90 days
- 0
- Latest release
- 0.16.8 (2026-05-14)
- Open issues
- 0
- Recent issues with at least one comment
- 30/30
Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.
Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.
Who it fits
- You want a long-lived agent that remembers past work and rewrites its own memory and context over time.
- You want one agent reachable from CLI, desktop, browser, mobile and messaging apps, with subagents, schedules and hooks.
- You want memory you can inspect: the README says context is tracked in git and can sync to your own repository.
- You only need to add memory to an agent you already use through MCP; a small MCP memory server is a lighter fit.
- You rely on older Letta tutorials or the V1 API server: the project restructured, and the V1 server is retired.
- You want everything to work without signing in: Letta Cloud is the default backend and some features need it.
How it differs
vs mcp-memory-service★ 2.1kLetta is a whole agent harness whose memory is part of the agent. MCP Memory Service is a memory add-on that plugs into any MCP client you already use, so pick it if you want to keep your current assistant.
vs LycheeMem★ 1.1kLycheeMem is a memory framework you embed in your own agent, with working, semantic and procedural stores. Letta ships the agent itself.
Compatibility
Quick start
Use cases
Alternatives
Featured in these lists
Comments
No comments yet. Be the first!