AgriciDaniel/claude-obsidian: A Trending AI-Powered Self-Organizing Knowledge Graph for Obsidian
The open-source Python repository AgriciDaniel/claude-obsidian has gone viral on GitHub, gaining over 2,600 stars in a single week. The tool integrates Claude Code with Obsidian to automatically read, link, and organize files into a self-organizing markdown knowledge graph. This project provides a practical, open-source alternative to proprietary knowledge management tools like Notion by leveraging local markdown files. It demonstrates the growing trend of using agentic AI tools to automate personal knowledge management (PKM) workflows. The tool is built on Andrej Karpathy's "LLM Wiki" pattern, which compiles raw documents into a persistent, structured wiki that Claude can query efficiently. Users retain full ownership of their data since the output consists of plain markdown files.
## BACKGROUND
Obsidian is a popular markdown-based note-taking application that visualizes notes as an interconnected graph. The "LLM Wiki" pattern, proposed by Andrej Karpathy, suggests compiling raw documents into a structured wiki once to avoid having LLMs re-read raw data repeatedly, using tools like Anthropic's Claude Code to query and maintain it.