Building a Personal Operating System with Obsidian and Claude

I've been building AI systems since the early 1990s. Expert systems at Boeing, two tours at Carnegie Mellon, what we used to call Knowledge Engineering. Extract what domain experts know, encode it into rules, deploy it. Thirty years later, the same fundamental problem is back, except now the tools are extraordinary.

At AIDC, I've built a network of AI agents. Custom GPTs for my CRO, CMO, PR, and Demand Gen functions, orchestrated through n8n workflows. They work. The gap was that these agents were disconnected from my personal knowledge base. My notes, strategic thinking, meeting captures, daily observations about the business and market.

That gap is what got me experimenting with Obsidian.

I had been using Reflect Notes. Great app, still recommend it. Fast, clean, excellent voice transcription. It's also a closed system. Encrypted cloud, proprietary format. Great for privacy, terrible for agent integration.

Obsidian stores everything as plain markdown files on your local filesystem. Just a folder of .md files. That single architectural decision changes everything when you're building an AI-augmented knowledge system.

Claude Code can read and write to any folder on your computer. Your vault is just a folder. So Claude gets full access to every note. Searching, creating, backlinking, organizing, extracting action items, all through natural language. Claude Cowork lets me schedule recurring tasks against the vault. A weekly review runs every Sunday morning, reads my daily notes, extracts decisions and action items, flags orphaned notes. Runs while I'm having coffee.

The real power is MCP, the Model Context Protocol. An Obsidian plugin turns your vault into an MCP server. My n8n workflows write files into the vault. My custom agents pull context from my notes before generating outputs. Everything converges on the same set of markdown files.

I'm calling it JARVIS. My Personal Operating System. I know, not the most creative name, but it's fun. I built a detailed integration guide covering the four paths to connect Obsidian with Claude, vault structure, agent skills, and the plugin stack I'm using. Linked below.

What we're assembling right now, personal knowledge bases connected to specialized AI agents through standardized protocols, is the closest thing I've seen to what we were dreaming about in those early Knowledge Engineering days. Except now it actually works, runs on a laptop, and costs less than a nice dinner.

Read the full Obsidian + Claude Integration Playbook

Dave Drach is the CEO and founder of AIDC (AI Data Co-op), providing governed, expert-curated datasets for enterprise AI through the Databricks Marketplace. Serial entrepreneur, Denver, Colorado.