Letta provides a platform for building memory‑first AI agents that retain context and continuously learn from experience. Developers use the Letta Code CLI and TypeScript SDK to create persistent, identity‑driven agents that can run locally or in the cloud, version their knowledge in a git‑backed repository, and be teleported across devices while integrating any LLM API.
Funding
$10M raised to dateRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
8OFounders
Product
Problem
Current large‑language‑model agents operate in stateless sessions, so they cannot retain knowledge across interactions, learn from their own experience, or improve over time. This limits their usefulness for long‑running tasks such as coding assistance, autonomous tool use, or personalized assistance.
Solution
Letta delivers a platform for memory‑first, stateful AI agents that persist their context and continuously learn from experience. The Letta Code runtime lets developers create agents that run locally or in the cloud, maintain a git‑backed memory repository, and execute “sleep‑time” compute to reflect and reorganize knowledge when idle. Agents can edit their own system prompts, load and version skills, and manage multi‑turn conversations through a TypeScript SDK. By storing learned information as token‑space representations rather than updating model weights, Letta agents keep their identity and expertise across model upgrades and can be teleported between devices without losing context. The platform also provides tools for agentic context engineering, skill learning, and scalable multi‑agent workflows.
Target Audience
Primary users are developers, data scientists, and AI product teams building autonomous assistants, coding agents, or enterprise automation solutions that require persistent, self‑improving AI. The platform also serves research labs and enterprises that need scalable, stateful agents for long‑term collaboration.
Features
- Persistent, identity‑driven agents with long‑term memory stored in a git‑versioned context repository
- Letta Code CLI and SDK (Node.js/TypeScript) for creating, resuming, and managing agents and conversations
- Built‑in “sleep‑time” compute that runs offline to reflect, consolidate, and improve token‑space memories
- Skill Learning pipeline that generates reusable skill files from past trajectories and feedback
- Multi‑model support and API‑key integration (OpenAI, Anthropic, OpenRouter, etc.) with Letta Auto model selection
- Remote deployment and teleportation of agents across machines while preserving full context
- Advanced context‑engineering tools (memory blocks, progressive disclosure, git‑based conflict resolution)