- CollectionAgent Runtime & Infrastructure
- Classificationagent-native
- MCPAvailable
- Agent SkillUnavailable
- InterfacesMCP · REST · SDK · API
What it does
Letta (evolved from the MemGPT research project at UC Berkeley) is a cloud platform for deploying and managing stateful agents at scale. Each agent deployed on Letta has persistent state, memory, conversation history, and reasoning traces stored in a model-agnostic format in the cloud — surviving restarts, model migrations, and provider changes.
The core abstraction: the agent is the deployment unit. Letta Cloud manages 24/7 uptime, auto-scaling, agent templates with versioning, and a visual debugging environment (Agent Development Environment, ADE).
Agent-facing interfaces
- Catalog protocols: MCP, REST, SDK, API
- MCP: ✅ Available
- Agent Skill: ⚠️ No official skill published by Letta yet.
- Onboarding type:
sdk
Protocol surface
| Interface | Detail |
|---|---|
| REST API | Full agent lifecycle: create, query, update, delete |
| Python SDK | pip install letta-client |
| TypeScript SDK | npm install @letta-ai/letta-client |
| MCP Server | Expose Letta agents as tools for other agents |
| ADE | Visual Agent Development Environment for debugging |
| Letta Cloud | Fully managed hosted deployment |
| Self-hosted | Open-source Docker deployment |