Descriptions:
David Ondrej walks through his personal setup for integrating Obsidian with the Hermes AI agent, creating what he describes as a “second brain” where every markdown file becomes a “living file” — directly accessible and actionable by AI agents rather than sitting inert on a hard drive. With 185 custom and pre-built Hermes skills already running in his own environment, Ondrej argues that routing agent memory through Obsidian’s visual, editable interface gives users genuine oversight of agent behavior and makes skill customization accessible without touching terminal config files.
The tutorial covers installing Hermes agent at the root level on a VPS using the Pi agent to handle all terminal operations via plain English instructions, configuring Obsidian Sync to keep files consistent across MacBook, phone, and server, and setting up OpenRouter as the inference provider. Ondrej also previews a private token-observability layer he’s building to track cost and usage per model and API key — highlighting a gap in current tooling around AI agent cost transparency.
The video is explicitly aimed at non-developers who want powerful AI automation without deep technical knowledge, framing Obsidian as the human-readable orchestration interface on top of Hermes. It is practically useful for anyone exploring personal AI agent setups, context engineering, or persistent memory architectures for productivity workflows.
📺 Source: David Ondrej · Published June 21, 2026
🏷️ Format: Tutorial Demo
