This is an early release preview. You may encounter bugs.
Framework
Letta logo
Unclaimed

memory

Letta

A server with CLI and API for stateful, self-improving agents with advanced memory.

B 78/100 GitHub score ? This grade is derived from GitHub signals — not user votes. Open for the full breakdown.
No votes yet

01 / About

About Letta.

Letta (formerly MemGPT) is a framework for building AI agents with advanced, persistent memory that learn and self-improve over time. It comes in two main forms: Letta Agent, a memory-equipped agent you run locally in the terminal, through a desktop app, or via channels like Slack, and the Letta Agent SDK (TypeScript), for embedding stateful agents into your own applications. Agents created with the SDK can run on Letta's hosted agent cloud (Constellation), fully locally on your machine, or against a self-hosted App Server. Agents are defined with a persona and knowledge about their human counterpart, hold long-running sessions, and stream responses back to the caller.

The CLI agent (Letta Code) supports skills and subagents and bundles pre-built ones for advanced memory and continual learning; it is model-agnostic, working across providers such as Anthropic and OpenAI. Note that this particular repository holds the legacy Letta server — the API server behind the V1 API and SDKs — and active development has moved to the Letta Agent repo, with self-hosting now done via the App Server. The previous-generation V1 client SDKs (TypeScript and Python) remain available, but the Agent SDK is the recommended path for new projects.

02 / Discussion CREDIBILITY-GATED

Discussion · 0

Reading is open to everyone. Only verified humans or builders at GitHub B+ can post or vote — every thread carries its author's credibility.

  • No discussions yet — start the first thread.

03 / Build

Build with Letta.

Browse the catalogue for frameworks, tools, and blueprints — each scored on real GitHub credibility.

Browse the catalogue