Letta Documentation | Letta Docs

# Letta > Letta agents learn from experience and improve with use. The Letta agent harness is fully open source and builds on our lab's latest research in AI memory and continual learning. Use Letta to create coding agents, personal assistants, AI coworkers, and stateful agents embedded in your own applications. Install the CLI: `npm install -g @letta-ai/letta-code` (Node.js 22.19+) Download the Letta app: [macOS](https://download.letta.com/mac/zip/arm64) · [Windows](https://download.letta.com/windows/nsis/x64) · [Linux](https://download.letta.com/linux/appImage/x64) Web app: https://chat.letta.com Repo: https://github.com/letta-ai/letta-code Letta's mission is to build machines that learn: agents that build memory, forge identity, and deepen knowledge from their own experience rather than living in short-lived, task-specific sessions. The core of Letta is defined by the [Context Constitution](https://github.com/letta-ai/context-constitution): the principles governing what a Letta agent places in its context window, in what order, at what level of detail, and for how long. Letta agents learn by actively managing their own context — creating durable token-space representations of their identity, memory, and continuity — rather than by updating model weights. Letta agents can extend their own capabilities through learning: acquiring new [skills](https://docs.letta.com/configuration/skills/index.md), system prompt learning (rewriting their own prompts as they accumulate experience), and even self-modifying the harness itself through [mods](https://docs.letta.com/configuration/mods/index.md) — adding tools, commands, and UI, changing how context is fed into the model, or adding custom model providers — to extend what they can do and customize the harness to how you work. Letta's main surfaces: - **The Letta Harness** (formerly Letta Code) — an open source harness designed for stateful, self-evolving agents. Build personal or shared agents that are p