Letta

Active

Country: United States (US)
Founded: 2023
Developer: Letta (AI lab, born from MemGPT at UC Berkeley's Sky Computing Lab)
Pricing: Freemium (free tier with own API keys, paid upgrades for quotas and access)
Open source: Yes
Platforms: macOS, Linux, Web, Cloud (e.g., Railway)

Indexes

APIs & MCP Servers

Overview

Letta is a platform for building stateful AI agents with persistent memory that learn and adapt from experience. It supports chat and voice interfaces, integrates with multiple LLMs, and enables self-hosting or cloud deployment. Targeted at developers and organizations for applications like personal assistants, customer support, and knowledge bases, it stands out with memory decoupling from model providers and tools for file system interaction.

Key Features

Persistent Memory - Agents retain memory, context, and personality across model switches and sessions.
Multi-LLM Support - Integrates with OpenAI, Anthropic, Mistral, and others via API keys.
Voice and Chat Interfaces - Supports both text-based chat and voice-controlled agents.
Letta Filesystem - Provides tools like grep, open, and semantic search for document interaction.
Self-Hosting - Fully self-hostable with privacy-focused design and no external data storage.
Desktop App (Letta Code) - Local runtime for coding agents with git-backed memory and skills.
Remote Agent Control - Teleport agents between machines while preserving memory.
Programmatic Tool Calling - API for agents to generate workflows with any LLM.

Pricing

Plan	Price	Includes
Free	Free	Use with own API keys, basic quotas, desktop/CLI/web access.
Paid Upgrade	Paid (details on site)	Centralized frontier model access, increased Letta Auto quotas.

Platforms & Requirements

Desktop app available for macOS; Linux supported via CLI and server mode. Web access through chat.letta.com. Cloud deployment on platforms like Railway; requires API keys for LLMs, no native Windows or mobile apps.

Integrations & Ecosystem

OpenAI (GPT/Codex)
Anthropic (Claude)
Mistral
xAI
Speech APIs for voice
Git for memory backing
Letta API for programmatic access
Railway for cloud deployment

Alternatives

App	Difference
MemGPT	Letta evolved from MemGPT, adding production runtime, voice support, and broader tooling.
LangChain	LangChain focuses on chains/workflows; Letta emphasizes stateful agents with persistent memory.
AutoGen	AutoGen specializes in multi-agent systems; Letta prioritizes single-agent memory and learning.
CrewAI	CrewAI for collaborative agent teams; Letta for individual persistent agents with file tools.

Reputation

Letta is recognized for advancing stateful AI agents from its MemGPT origins, praised for memory persistence and model flexibility. Users appreciate open-source self-hosting and tools like Filesystem for document handling. Some note dependency on external LLM API costs and early-stage limitations in multi-agent support.

Sources (8)