Hermes AI Agent: Windows setup and task automation

Learn to set up Hermes on Windows with a local model, manage YouTube replies, and schedule reviews. The demonstrated model needs about 18 GB of storage.

Player not loading? Watch on YouTube

Kevin Stratvert walks through installing Hermes on Windows and connecting it to a local LLM. He describes Hermes as free and open source, with Mac and Linux support as well. The tutorial uses the Windows desktop installer and automatic model setup. The recommended model, identified in the transcript as Qwen 3.8 27B, requires about 18 GB of storage and sufficient memory; the speaker gives no specific RAM requirement.

The first task creates a desktop folder and a document with automation ideas. Stratvert then connects Microsoft Edge through the browser command and opens an already signed-in YouTube Studio session. He asks Hermes to review five recent comments and draft replies without publishing them. A separate approval prompt authorizes one reply to be posted. This demonstrates computer and browser actions with local AI inference, rather than establishing that the website workflow works offline.

The tutorial also tests saved response preferences in a new conversation and turns the comment workflow into a reusable skill. A scheduled job invokes that skill daily at 9:00 a.m. to prepare replies for review. Stratvert notes that Hermes must remain running in the background and the browser session must stay accessible. He closes with an overview of connecting WhatsApp or Telegram to send requests from a phone.