Hermes Agent setup with Ollama and Obsidian

Tina walks through her Hermes Agent setup, starting with hardware and the macOS desktop app. She uses an M4 Mac Studio with 64 GB of RAM, but also discusses a VPS, a spare computer and a personal machine. She recommends Docker for the personal-machine option. The tutorial describes terminal installation for Mac and Windows as another route.

The demonstrations cover web search, messaging integrations and reusable skills. Tina turns a business-idea evaluation into a skill, then saves research to an Obsidian vault. Scheduled tasks produce daily briefings, compile notes and check her running systems. These are examples from her own configuration.

Memory gets a separate explanation: brief core files hold setup and user information, while session search retrieves archived conversations from a SQLite database. Tina adds Honcho for inferred patterns across interactions and uses Obsidian to store research and project plans that Hermes can retrieve later.

For those who want to run models locally, she demonstrates Ollama's Hermes launch option and explains her preference for llama.cpp. She attributes greater control and speed to that choice without presenting benchmarks. In this section, she identifies her model as Qwen 3.6 35B. The final demonstration uses Discord to request and queue a Pomodoro app build, tests its timer and checks the completion record in Obsidian. It introduces her multi-agent workflow rather than providing its full setup.