Letta Code is a memory-first coding harness designed for developers seeking persistent, evolving AI coding assistants. It addresses the limitations of stateless, session-based coding agents by providing a long-lived, portable agent that learns and retains context across interactions. This enables a more collaborative and efficient development workflow, akin to working with a continuously learning junior developer or mentee.

How It Works

Letta Code employs a philosophy centered on persistent agents rather than ephemeral sessions. Each interaction is tied to a singular agent that accumulates knowledge and improves over time. This agent-based approach contrasts with traditional session-based models where context is limited to the current conversation. Letta Code's core mechanism involves initializing and managing agent memory, allowing for explicit guidance via /remember commands and enabling the agent to learn new capabilities through a /skill command or by integrating reusable modules from a .skills directory.

Quick Start & Requirements

Installation is available via npm: npm install -g @letta-ai/letta-code. After installation, navigate to your project directory and run the letta command. Community-maintained packages are available on the Arch User Repository (AUR) for Arch Linux users (yay -S letta-code). Letta Code connects to the Letta Developer Platform (offering a free tier) via OAuth or a LETTA_API_KEY, or to a self-hosted Letta server using LETTA_BASE_URL. Initialization requires the /init command. Further details are available on the official docs page.

Highlighted Details

• Model Portability: Designed to work across multiple large language models, including Claude Sonnet/Opus, GPT-5, Gemini 3 Pro, and GLM-4.6.
• Persistent Memory: Agents retain learned information and context across sessions, unlike typical stateless coding assistants.
• Skill Learning: Supports both pre-defined skills in .skills directories and dynamic learning of new skills directly from the agent's interaction history.

Maintenance & Community

The project is marked as "Made with 💜 in San Francisco." Specific details regarding core maintainers, active community channels (like Discord/Slack), or a public roadmap are not detailed in the provided README.

Licensing & Compatibility

The license type and any associated restrictions for commercial use or closed-source integration are not specified in the README. This absence requires further investigation before adoption.

Limitations & Caveats

The README does not explicitly detail limitations, known bugs, or alpha/beta status. The reliance on the Letta API or a self-hosted server implies a dependency on external infrastructure or setup. The lack of explicit licensing information presents a significant adoption blocker for commercial or sensitive projects.