Hermes Agent setup with Hostinger, OpenRouter and Telegram

This beginner tutorial sets up Hermes Agent through Hostinger's one-click VPS template, then connects an OpenRouter API key and selects DeepSeek V4 Flash. The presenter describes Hermes as free software from Nous Research with an MIT license. The demonstrated self-hosted setup still uses a paid server and an external model provider.

The guide explains the choice between running the agent on a computer and keeping it on a server. The desktop route stops when the laptop sleeps, according to the presenter. He installs the macOS app and connects it to the server through a remote gateway rather than creating a separate agent. He also describes the Windows flow.

Telegram setup uses a QR code, followed by a gateway restart and a home channel for scheduled results. The tutorial then tests a saved writing preference in a new session. In another example, Hermes creates a reusable skill to check YouTube for AI productivity trends and schedules it every four hours, with reports sent to Telegram.

The final sections cover model spending limits, usage logs, gateway and application restarts, and parallel sub-agents. The presenter also discusses prompt injection risks when an AI agent can browse, execute code and access files. His reported model costs describe this recording's usage, rather than a guaranteed running budget.