Hermes setup with Ollama, VPS hosting and OpenClaw

This tutorial follows the speaker's Hermes setup, from installation and importing OpenClaw settings to Telegram access and scheduled workflows. It covers the terminal interface, a browser dashboard for sessions and configuration, and reusable skills. The speaker notes that the dashboard shown lacks a web chat interface.

For a local LLM setup, the walkthrough uses Ollama on a MacBook and the command ollama launch hermes. The presenter switches models after his first choice fails to provide tool calling. Hardware limits which models can run effectively. A separate self-hosted deployment uses OpenRouter Spawn to create a DigitalOcean VPS, with OpenRouter supplying model access. The routing discussion covers task-specific model choices and fallback providers.

Practical examples include daily Apple Health sleep reports through the presenter's Health Data AI Analyzer app, Threads post analytics, and Gmail and Google Calendar access through enabled APIs and desktop OAuth credentials. The final setup gives Hermes coordination duties and delegates longer workflows to OpenClaw. Both use markdown files on a NAS, which the speaker also edits through Obsidian. His preference for Hermes memory and OpenClaw's handling of complex tasks reflects his own experience rather than a controlled comparison.