Summary
Youri van Hofwegen walks through a complete deployment of Hermes, an open-source AI agent from Nouse Research that has accumulated over 181,000 GitHub stars. Unlike conversational AI assistants, Hermes runs on a persistent schedule — sending morning briefings, pulling weekly metrics, or processing repetitive tasks and returning finished results without the user being present. The tutorial covers everything from provisioning a Hostinger KVM2 VPS (the recommended host for always-on operation) through first conversation with the agent via Telegram.
Setup involves a web terminal session to install and configure Hermes, connecting it to Open Router — which unlocks over 400 AI models from a single API key — and running through the full setup wizard to configure messaging integration. Van Hofwegen walks through creating a Telegram bot via BotFather, obtaining a user ID via userinfobot, designating a home channel for scheduled notifications, and creating a Telegram group with topics for organizing different agent workflows (social media, analytics, personal tasks).
Practical cost management is also covered: Hermes includes auto-routing that assigns cheap models to lightweight tasks (a few cents per run) and stronger models to complex reasoning (a couple of dollars), with optional per-key spending caps in Open Router to prevent runaway costs. A key operational note: the Hermes gateway process must remain running for the agent to receive messages and fire scheduled jobs — if it stops, the bot goes silent and scheduled tasks don’t execute. Total time from deploy to first response is approximately ten minutes.
📺 Source: Youri van Hofwegen · Published July 03, 2026
🏷️ Format: Tutorial Demo
