Hermes Agent setup: VPS, Telegram and GitHub backups

This tutorial sets up Hermes Agent as a self-hosted personal assistant on a Hostinger VPS. The speaker chooses Ubuntu 24.04 LTS and a one-click Docker deployment, then authenticates through OpenAI Codex with a ChatGPT account and selects GPT 5.5. This walkthrough uses a remote inference provider; it does not demonstrate how to run models locally.

Telegram setup covers creating a bot through BotFather, adding its token and restricting access to the speaker's user ID. The first connection fails because the gateway is stopped. The speaker asks Hermes to investigate, then confirms that messages work after it restarts the gateway.

The course explains memory files, reusable skills, soul.md personality settings and scheduled tasks. The speaker describes improvement as saving useful experience and correcting skills, rather than a process that needs no supervision. Cron jobs start isolated sessions, so their prompts need enough context to stand alone. Another exercise installs the HyperFrames skill and generates a short video from soul.md.

A GitHub backup exercise creates a private repository, excludes secrets and builds a nightly sync skill. The schedule checks Central Time hourly to handle daylight saving changes. The speaker also compares Hermes with Claude Code and OpenClaw, keeping Claude Code as his desk-based coding assistant. Later sections cover scoped credentials, firewalls and separate containers for agents with different roles. Telegram provides less visibility into context and fewer commands than the terminal.