Yggdrasil is a powerful tool designed to enhance AI agents by providing them with persistent semantic memory, allowing repositories to become self-aware. This innovative system structures intent, constraints, and architecture into a graph that persists beyond context windows, addressing the limitations of ephemeral memory in conversation and execution of code.

Key Features

• Persistent Semantic Memory Graph: Yggdrasil maintains a structured understanding of your project's system intent, rules, and boundaries using Markdown and YAML, all organized within a .yggdrasil/ directory.

• Deterministic Context Assembly: When an AI agent interacts with your code, Yggdrasil assembles precise context packages tailored to the specific component being modified. This eliminates the need for the agent to guess, thereby enhancing the quality of output.

• Seamless Integration: Yggdrasil does not alter existing workflows. It functions as an ambient infrastructure that allows AI agents to read the memory graph, write code, and automatically update the graph based on new decisions.

Supported Platforms

Yggdrasil is designed to streamline the work of AI agents across various platforms, including:

• Cursor
• Claude Code
• GitHub Copilot
• Cline / RooCode
• Codex
• Windsurf
• Aider
• Gemini CLI
• Amp

Not a Code Generator or Invasive Tool

This project is not a code generator; rather, it optimizes your existing AI agent's ability to generate high-quality code. Unlike traditional documentation, the memory graph is updated naturally by the agent during tasks, eliminating manual documentation overhead. Importantly, Yggdrasil is non-invasive; users can easily remove the .yggdrasil/ folder if needed, restoring the project to its original state.

Learn More

For comprehensive documentation on specifications and architectural concepts, visit the Yggdrasil documentation. By leveraging Yggdrasil, repositories gain the capability to understand their own architecture, ensuring a cohesive and informed development experience.