Hermes Agent goes a step beyond a chatbot – it’s an autonomous AI agent that can browse the web, run commands, schedule tasks, and remember context across sessions. In this guide I’ll walk through cloning an Ubuntu VM from your Proxmox template, installing Hermes Agent, connecting it to a free-tier model via the Nous Portal, and wiring it up to a Telegram bot so you can talk to your agent from your phone – all running on your own hardware with no GPU and no subscription required.
🎥 Watch the Video Tutorial
💡 What Is Hermes Agent?
A typical chatbot responds to messages – you ask, it answers, conversation ends. An agent is different: it can take actions on your behalf. Hermes Agent, from Nous Research, can browse the web for information, execute commands, run scheduled tasks, and maintain persistent memory across conversations. It behaves less like a Q&A box and more like an assistant that actually does things.
The setup in this guide uses the Nous Portal free tier for the underlying model – no local LLM, no GPU, and no subscription cost. Hermes runs as a lightweight service on an Ubuntu VM, and you interact with it through Telegram from anywhere.
🛠 What You’ll Need
- A cloned Ubuntu Server VM from your Proxmox template
- Hermes Agent (Nous Research) – free, open source
- A Nous Portal account for free-tier model access
- A Telegram account and the Telegram app on your phone
📋 Step-by-Step Setup
1. Clone the VM from your template
In the Proxmox web UI, right-click your Ubuntu Server template and select Clone. Choose Full Clone and give it a name that reflects its purpose – e.g. hermes-agent. Start the cloned VM and update its hostname to match:
sudo hostnamectl set-hostname hermes-agent
SSH into the new VM to continue.
2. Install Hermes Agent
Download and run the Hermes Agent installer from Nous Research’s GitHub. The installer walks you through a setup wizard – work through each prompt as it appears.
3. Run through the setup wizard
The setup wizard configures Hermes for first use. Key prompts to expect:
- Confirm the install directory and any required dependencies
- Choose your model provider – select the Nous Portal option for free-tier access
- Generate or paste your Nous Portal API credentials
- Set any local storage paths for memory and logs
4. Connect to the Nous Portal free tier
Sign up for a Nous Portal account if you haven’t already. The free tier gives you access to a hosted model without running anything locally – no GPU required, and Hermes routes its requests through the portal. During the setup wizard, paste your Nous Portal API key when prompted. This is what gives Hermes its reasoning capability without needing Ollama or any local model.
5. Create a Telegram bot with BotFather
Hermes connects to Telegram through a bot account. To create one:
- Open Telegram and search for @BotFather – Telegram’s official bot creation account
- Send
/newbotand follow the prompts to name your bot - BotFather returns a bot token – copy this, you’ll need it for Hermes
Paste the bot token into Hermes’ configuration where prompted during setup, or into the relevant config file if configuring after the wizard.
6. Restrict bot access to your Telegram user ID
By default a Telegram bot will respond to anyone who messages it. Since Hermes can run commands and browse the web on your behalf, restrict access to your own Telegram user ID – and anyone else you explicitly trust:
- Find your Telegram user ID by messaging @userinfobot on Telegram – it replies with your numeric ID
- Add your user ID to Hermes’ allowed users configuration
7. Access the Hermes web dashboard remotely
Hermes includes a web dashboard for monitoring agent activity, reviewing logs, and managing configuration – useful even though the VM itself is headless. Since the VM has no desktop environment, access the dashboard from your browser using the VM’s IP address and the port Hermes listens on (check the setup wizard output for the exact port).
8. Explore scheduled tasks, skills, and memory
With Hermes running and connected to Telegram, explore what it can do beyond simple chat:
- Scheduled tasks – configure Hermes to run actions automatically on a schedule, similar to a cron job but driven by the agent
- Skills – modular capabilities Hermes can use, such as web browsing or command execution
- Persistent memory – Hermes remembers context across separate conversations, so you don’t need to re-explain things every session
Message your bot on Telegram to confirm everything is working – you should get a response routed through the Nous Portal model, with Hermes able to take follow-up actions based on the conversation.
✅ Conclusion
You now have a self-hosted autonomous AI agent running in your home lab – accessible from your phone via Telegram, powered by a free-tier cloud model, with no GPU and no subscription. Hermes can browse the web, execute commands, run scheduled tasks, and remember context across sessions, going well beyond what a typical chatbot offers. From here, explore the available skills and start building scheduled tasks that handle real recurring jobs in your lab.
Previous guide: How to Create a Reusable Ubuntu 26.04 Server Template in Proxmox
📺 Watch the full video guide here: https://youtu.be/glZok_TSpco
If you found this helpful, like and subscribe to IT HomeLab Online on YouTube for more tutorials.
☕ Support the channel: Patreon · Buy Me a Coffee