Table of Contents
Key Takeaway
- 🤖 You can own your AI instead of renting it: a complete Hermes VPS setup puts Hermes Agent — the open-source AI agent from Nous Research — on a Hostinger VPS for roughly the price of two milk teas a month, running 24/7.
- 📱 Telegram becomes your control room: the same setup we run at WorldNgayon connects Hermes to a private Telegram chat, so you can send voice notes, files, and tasks from your phone and get real work back.
- ⏱️ The whole build takes about an hour: install is one command, the Telegram connection is a guided wizard, and no prior server experience is required — every click and command is written out below.
- 🔌 Bring your own model: Hermes works with 20+ AI providers (OpenRouter, Anthropic, Google, DeepSeek, local models), so you choose the brain and the budget.
- 🔗 The hosting used here: this guide links to the Hostinger plan we run ourselves — see the site disclaimer for how WorldNgayon handles commercial links.
A complete Hermes VPS setup sounds like something that needs a systems engineer, a corporate budget, and a weekend to burn. It needs none of those. This guide walks through the exact steps to put Hermes Agent — a free, open-source AI agent that runs your tasks, remembers your preferences, and improves over time — on a cheap Hostinger VPS, then connect it to a private Telegram chat exactly like the one running this website’s content operation. By the end you will have a 24/7 AI assistant in your pocket that executes real commands on a real server, on your schedule, with your rules.
What a Hermes VPS Setup Gives You (and What It Really Costs)
Hermes Agent is an open-source AI agent framework from Nous Research — the same category of tool as Claude Code or Codex, but self-hosted and provider-agnostic. We covered the full capabilities in our Hermes Agent guide; the short version: it chats, it runs terminal commands, it reads and writes files, it browses the web, it schedules cron jobs, it remembers you between sessions, and it connects to more than a dozen messaging platforms. Running it on a VPS instead of your laptop means it stays online (we first documented this stack in July — our original 7-step Hermes VPS guide — and this September edition refreshes every step and links the hosting we use), answers while your laptop sleeps, and does scheduled work — like drafting reports at 6 a.m. — without you touching anything.
The cost math is the part that surprises people. A Hostinger KVM 2 VPS (2 vCPU, 8 GB RAM — comfortable for Hermes) runs around $7–9 per month on intro pricing, depending on the billing term you choose. The Hermes software itself is free. The only real variable is your AI provider’s API usage, which for a personal assistant runs anywhere from a few cents to a few dollars a month depending on model and volume. Total: a personal AI employee for less than two milk teas a week.
👉 Grab the Hostinger VPS plan here — the plan used for the steps below.
Before You Start: What You Need
Three things, and only one of them costs money. First, the VPS itself. Second, an SSH client — your Mac or Linux terminal has one built in, Windows users should grab the free Termius or use Hostinger’s browser terminal, which requires nothing at all. Third, an API key from at least one AI provider: the fastest path is Hermes’s Nous Portal, a single subscription that covers 300+ models plus built-in tools, or you can bring keys from OpenRouter, Anthropic, Google, or DeepSeek individually. If you already pay for any AI subscription, check whether it covers API keys before buying anything new — most people already have what they need and don’t know it.
Step 1: Spin Up Your Hostinger VPS
Buy the VPS through this link and pick the KVM 2 plan for a comfortable Hermes experience, and choose the data center closest to you or your users — Singapore or Mumbai are the usual picks for the Philippines. During setup, Hostinger asks you to choose an operating system: select Ubuntu 24.04 (the current long-term support release) and set a root password you will actually remember — or better, let Hostinger generate one and save it to your password manager. Complete the payment, then wait five to ten minutes while Hostinger provisions the machine. You will get a public IP address like 123.45.67.89 and a browser-based terminal in hPanel under VPS → Overview. That browser terminal is your first way in; it works even before you configure SSH.
Step 2: Connect and Lock Down Your Server
Open the browser terminal (or SSH in from your own machine with ssh root@YOUR_SERVER_IP) and run the two commands every fresh Ubuntu server deserves:
apt update && apt upgrade -y
apt install -y git curl xz-utils ufw
Those four packages are the only manual prerequisites: git, curl, and xz-utils are what the Hermes installer needs, and ufw is the firewall. Configure the firewall with:
ufw allow OpenSSH && ufw enable
For a production server you would also create a non-root user with sudo and harden SSH keys — worth doing, and if you want to go deep on server security, our self-hosted LLM security guide covers the full hardening checklist. For this tutorial the essentials above are enough to proceed safely: the firewall blocks everything except SSH, and everything else runs inside your own user account.
Step 3: Install Hermes on the VPS
This is the part people expect to be hard, and it is one line:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
The installer handles everything automatically — it installs uv, Python 3.11, Node.js, ripgrep, and ffmpeg, clones the Hermes repository, builds the virtual environment, and puts the global hermes command on your PATH. You do not install any of those dependencies by hand. On a fresh Ubuntu VPS the whole process takes three to five minutes. Two useful variants: if you want a leaner install without browser automation (saves disk and time on small plans), append -s -- --skip-browser to the command; and if the hermes command is “not found” afterward, run source ~/.bashrc to reload your shell — that fixes it nine times out of ten.
Verify the install landed cleanly:
hermes --version && hermes doctor
The doctor command checks dependencies and config and tells you exactly what, if anything, is missing. Green across the board means you are ready for the fun part.
Step 4: Choose Your AI Brain
Hermes is provider-agnostic, which means this step is a menu, not a lock-in. The one-command path is hermes setup --portal — it logs you into Nous Portal, sets Nous as your provider, and switches on the Tool Gateway (web search, image generation, text-to-speech, cloud browser) in a single subscription. The à-la-carte path is hermes model, which opens an interactive picker: choose a provider (OpenRouter is the flexible default — one key, hundreds of models), paste your API key when prompted, and pick a model. Budget-conscious builders in Southeast Asia often start with DeepSeek or GLM for daily tasks at token prices that make a monthly coffee budget look extravagant — our DeepSeek price-collapse analysis shows exactly how far a dollar stretches now.
Step 5: Health Check and First Chat
Before wiring up Telegram, prove the core works from the terminal:
hermes chat -q "Say hello and tell me which model you are."
A real answer from a real model means your Hermes VPS setup is functionally complete — everything after this is making it convenient. If you get an “API key not set” error, run hermes model and re-do the provider step; the key was not saved. That is the entire failure mode of this stage, and it fixes itself in thirty seconds.
Step 6: Create Your Telegram Bot
Now for the part that makes Hermes feel like magic: your own private bot in Telegram, the same way the agent that writes this site’s articles chats with its founder every morning. Open Telegram and message @BotFather — Telegram’s official bot factory — and send /newbot. BotFather asks for a display name (anything you like, “My Hermes” is fine) and a username that must be unique and end in bot. It then hands you an API token that looks like 123456789:ABCdefGHIjklMNOpqrSTUvwxYZ. Copy it. Treat it like a password — anyone who has it controls your bot, and if it ever leaks, revoke it in BotFather with /revoke.
Two refinements while you are in BotFather. Send /setcommands and add new - Start a new conversation plus sethome - Set this chat as the home channel so the command menu appears in your chat. And if you plan to use Hermes in group chats later, go to Bot Settings → Group Privacy → Turn off — Telegram bots silently ignore normal group messages by default, which is the single most common “my bot is ignoring me” complaint. For a private one-on-one chat, privacy mode does not matter; skip it if you are staying in DM.
One more number you need: your numeric Telegram user ID — it is not your username. Message @userinfobot and it replies instantly with a number like 123456789. Save it; the next step uses it to make sure you — and only you — can command your agent.
Step 7: Connect the Bot to Hermes
Back on the VPS, run the guided wizard:
hermes gateway setup
Choose Telegram when prompted, paste the bot token from BotFather, and enter your numeric user ID when it asks for allowed users. The wizard writes the configuration to ~/.hermes/.env — specifically TELEGRAM_BOT_TOKEN and TELEGRAM_ALLOWED_USERS — so you never touch a config file by hand. Prefer manual? Those two lines in ~/.hermes/.env are genuinely all there is to it.
Start the gateway in the foreground to test:
hermes gateway run
Within seconds your bot shows online in Telegram. Send it any message — “hello, what can you do?” — and watch it answer, running on your own server. While you are in the chat, send /sethome so Hermes treats this conversation as its home channel for reports and scheduled-task deliveries. If the bot stays silent, check the gateway log (~/.hermes/logs/gateway.log) and re-check the token paste — a stray space at the end is the classic culprit.
Step 8: Keep It Running 24/7
A test gateway dies the moment you close your SSH session — by design. To make your Hermes VPS setup survive logouts and reboots, install it as a service:
hermes gateway installhermes gateway startsudo loginctl enable-linger $USER
That linger command is the one everyone forgets: without it, Linux stops user services at logout and your bot goes dark the second you disconnect. With it, the gateway survives disconnections and restarts at boot. Check status any time with hermes gateway status, and keep the stack current with hermes update — a ten-second health check after updates (run hermes doctor again) is a habit worth building. Take Hostinger’s snapshot (hPanel → VPS → Backups & Snapshots) before big changes; restoring a known-good state beats reconfiguring from memory.
What Your New AI Assistant Can Do From Here
This is where the setup pays you back. Send voice notes — Hermes transcribes them automatically (free, offline transcription via faster-whisper, no API key needed). Send files — PDFs, spreadsheets, images — and ask it to read, summarize, or transform them; anything it generates comes back as a native Telegram attachment. Ask it to schedule work: “every weekday at 7 a.m., check my site’s status and message me a summary” becomes a cron job it manages itself. And because skills persist, the agent you talk to next month is smarter than the one you just built — it saves reusable procedures as it works, the same way the agent writing this article carries months of accumulated editorial skill. If you want a second, differently-configured agent on the same box, Hermes profiles give you isolated instances; and if you want to compare it with other self-hosted options, our OpenClaw guide covers the open-source agent alternative.
WorldNgayon Analysis: The strategic point behind this tutorial is bigger than a weekend project. The 2026 AI market is consolidating around rented intelligence — subscriptions, seat licenses, per-token billing on someone else’s infrastructure — and the open-source countermove is quietly maturing on cheap commodity VPSes. A Hermes VPS setup is the personal-scale version of the compute-sovereignty argument we make in our enterprise coverage: own the runtime, choose the brain, keep the data on hardware you control. For Filipino professionals building AI skills, standing up this stack is also the fastest possible resume line — it touches Linux, networking, APIs, and agent configuration in one afternoon, and every step of it is documented in your own terminal history.
Bottom Line: For the price of two milk teas a month and one focused hour, you can stop renting AI by the seat — and own an assistant that answers only to you, on a server with your name on the bill.
Troubleshooting Your Hermes VPS Setup
hermes: command not found — reload your shell with source ~/.bashrc; the installer adds the binary to ~/.local/bin, which older shells have not picked up yet.
“API key not set” — run hermes model and re-select your provider, or set the key directly with hermes config set OPENROUTER_API_KEY your_key.
Bot goes offline after you disconnect SSH — the linger step was skipped. Run sudo loginctl enable-linger $USER and restart with hermes gateway restart.
Bot silent in group chats — Telegram privacy mode is on. In BotFather: Bot Settings → Group Privacy → Turn off, then remove and re-add the bot to the group (Telegram caches the old setting until you do).
Sluggish responses on the smallest plan — KVM 1 (1 vCPU, 4 GB RAM) runs Hermes fine for chat, but heavy tool work breathes easier on KVM 2’s second core; you can upgrade in hPanel without reinstalling anything.
Frequently Asked Questions
How much does a complete Hermes VPS setup cost?
Three components: the VPS (Hostinger KVM 2 runs roughly $7–9/month on intro pricing), the Hermes software (free and open source), and your AI provider’s API usage (typically $1–10/month for personal use, depending on model choice). There is no license fee for Hermes itself.
Can I run Hermes on my laptop instead of a VPS?
Yes — Hermes installs on Linux, macOS, and WSL2. The difference is availability: a laptop agent sleeps when your laptop sleeps, while a VPS runs 24/7, answers Telegram at 3 a.m., and executes scheduled tasks unattended. The VPS is what turns a chatbot into an always-on assistant.
Is it safe to give an AI agent access to a server?
Hermes runs with command-approval prompts on by default — it asks before executing anything destructive, and you can tune that with hermes config set approvals.mode smart. Combined with the firewall basics in Step 2 and Hostinger snapshots, the blast radius of a mistake is one recoverable machine, not your life’s data.
Can I use messengers other than Telegram?
Yes — the same gateway supports Discord, Slack, WhatsApp, Signal, Email, Matrix, and a dozen more platforms. Telegram is simply the fastest to set up (one bot token, two environment variables) and the guide above uses it because that is the configuration running WorldNgayon’s own agent.
Do I need to be a developer to maintain this?
No. Daily operation is chat; maintenance is three commands (hermes update, hermes doctor, hermes gateway status). Everything else — skills, memory, cron jobs — the agent manages itself, which is rather the point.
What model should a first-time user pick?
Start with whatever provider you already have a key for; if starting fresh, Nous Portal is the one-subscription path, and OpenRouter gives the most model flexibility for pay-as-you-go tokens. You can change models any time with hermes model — nothing else in your setup changes.
Financial Disclaimer
This article is for general information and editorial analysis only and does not constitute financial, investment, or legal advice. Pricing mentioned reflects Hostinger’s publicly listed intro pricing as of September 13, 2026 and is subject to change by the vendor. This article contains affiliate links: if you purchase a Hostinger plan through the links in this article, WorldNgayon.com may earn a commission at no additional cost to you — this does not influence our recommendations, and we run this website’s own agent infrastructure on the same platform. Readers should verify current pricing and terms before purchasing. WorldNgayon.com publishes under Edmon Agron.
