Letta
Letta (formerly MemGPT) builds stateful AI agents with persistent memory that runs as a service. The Letta REST API creates, configures, and messages agents whose memory - core context blocks and archival vector memory - survives across sessions, and manages the tools, data sources, identities, and multi-agent groups those agents use. It is offered as a managed Letta Cloud API and as an open-source, self-hostable server with the same OpenAPI-documented interface.
APIs
Letta Agents API
Create, list, search, count, retrieve, modify, delete, import, and export stateful agents - each with its own persistent memory, tools, sources, and model configuration. Agents ...
Letta Messages API
Send synchronous, asynchronous, and streaming messages to an agent, list and search message history, cancel an in-flight run, preview the raw model payload, and reset or summari...
Letta Memory Blocks API
Manage core-memory blocks - the labeled, always-in-context text that gives an agent its persona, human, and task memory. Create, list, retrieve, modify, and delete standalone bl...
Letta Archival Memory API
Manage archives - an agent's out-of-context, searchable long-term memory - and the passages stored inside them. Create and list archives, attach or detach an archive to an agent...
Letta Tools API
Create, list, search, count, upsert, modify, delete, and directly run tools (Python functions) that agents call. Includes registering the built-in base tool set and executing a ...
Letta MCP Servers API
Register, connect, test, refresh, update, and delete remote MCP (Model Context Protocol) servers, list the tools each server exposes, add an individual MCP tool to an agent's to...
Letta Sources and Files API
Create and manage data sources (also exposed as folders) that hold uploaded files for agent grounding and retrieval - upload files, list and delete files and their extracted pas...
Letta Identities API
Model the end users of a multi-tenant Letta deployment as identities. Create, list, count, retrieve, modify, upsert, and delete identities; set custom identity properties; and a...
Letta Multi-Agent Groups API
Create, list, retrieve, modify, and delete multi-agent groups that coordinate several agents (round-robin, supervisor, and other patterns) around shared memory blocks. Attach or...
Letta Models and Providers API
List available LLM and embedding models, and register, check, list, retrieve, modify, delete, and refresh bring-your-own-key (BYOK) model providers (OpenAI, Anthropic, Google, A...
Letta Runs, Jobs and Steps API
Inspect the asynchronous execution history behind agent messages. List and retrieve runs and their messages, metrics, usage, trace, and stream; list and cancel background jobs; ...
Letta Agent Templates API
Cloud-only API for versioning agent configurations as templates - create, save, roll back, rename, and delete template versions, snapshot and restore a template from an agent fi...
Letta Chat Completions API
OpenAI-compatible chat completions endpoint that routes requests to a Letta agent, plus a companion voice-beta chat completions endpoint tuned for low-latency voice agent integr...
Collections
Letta API
OPENPricing Plans
Rate Limits
FinOps
Letta Ai Finops
FINOPSEvent Specifications
Letta Agent Message Streaming (HTTP + SSE)
AsyncAPI 2.6 description of Letta's **agent message streaming** surface. Letta does not publish a WebSocket API. The only asynchronous / event-style transport documented in Lett...
ASYNCAPI