Letta Stateful Agents

Build stateful agents with persistent memory using Letta (formerly MemGPT). Configures memory blocks, archival memory, tool execution, and self-editing memory loops that survive across sessions and process restarts.

This skill helps you build long-running agents with Letta's memory hierarchy. It configures core memory blocks, archival memory for long-term recall, tool definitions, and the self-editing memory loop where the agent rewrites its own context. Covers Letta Cloud deployment, local self-hosting with PostgreSQL, multi-agent shared memory, and patterns for stateful customer support, research assistants, and long-running workflows.

When to use

Use when building agents that need persistent memory across sessions, multi-turn assistants where the agent should remember user preferences, research agents accumulating knowledge over time, or any workflow where stateless agents repeatedly relearn context.

Examples

Stateful support agent

Build a customer support agent with persistent memory

Create a Letta agent for customer support that remembers each user's account history, preferences, and past issues across sessions, with tools for ticket lookup

Memory block design

Design core memory blocks for a research assistant

Design Letta core memory blocks for a research assistant that tracks current project, sources reviewed, open questions, and writing voice across long sessions