I use Hermes as a personal AI assistant that I can reach from my phone. It runs on a VPS, works with my business files, and brings important actions back to me for approval.
Start here
The first guide walks through the setup I use: an always-on VPS, Hermes, an AI model, Telegram, a durable Markdown workspace, GitHub, and a small set of connected tools.
It also explains the part most setup guides skip: where the useful files live, how memory works, and what the assistant should never do without asking.
What I mean by an AI assistant
I am not trying to build a chatbot that only answers questions. I want an assistant that knows where the work lives and can help move it forward.
That means it can research, organize files, prepare website changes, help package videos, and monitor recurring work. It still needs clear limits. Publishing, spending money, deleting data, and difficult-to-reverse decisions come back to me.
What I am testing
- Keeping useful business context in durable files instead of old chat threads
- Using Telegram as the control room while the work happens on the VPS
- Giving tools the smallest permissions needed for the job
- Using Git history, previews, and approval gates before changes go live
- Adding recurring jobs only after the manual workflow works
The goal is not maximum autonomy. It is a useful assistant that remembers the business without quietly taking over the business.