How to install and run Hermes Agent on a Hostinger VPS (with Telegram, skills, and cron jobs)

If you want your own always-on AI agent that lives on your server, talks to you through Telegram, and quietly does work in the background, Hermes Agent is a great option. This guide walks you through setting it up on a Hostinger VPS, connecting it to OpenRouter and Telegram, and then giving it skills, memory, and scheduled jobs so it can actually do useful work for you.

What Hermes Agent is and why run it on a VPS

Hermes Agent is a free, open-source AI agent from Nous Research. Think of it as a personal version of ChatGPT or Claude that runs on your own infrastructure instead of someone else’s cloud interface.

Unlike a typical chatbot, Hermes is designed to:

• Run continuously on a server
• Talk to you through apps like Telegram, Discord, or Slack
• Use tools (like a browser) and write its own reusable skills
• Remember how you work over time and use that memory in new tasks
• Schedule and run jobs even when you’re offline

Running Hermes on a VPS like Hostinger gives you a stable, always-on environment with a public endpoint, so your agent can stay online 24/7 and respond from your phone or laptop anywhere.

Choosing the right Hostinger VPS plan

Hostinger offers several KVM VPS tiers. For Hermes Agent, the recommended plan is KVM 2. It includes:

• 2 CPU cores
• 8 GB RAM
• 100 GB storage

Hermes runs inside Docker and, in this setup, will also use a browser for web research. That combination needs more resources than the cheapest plan if you want it to stay responsive and not freeze when browsing or running multiple tools.

On the Hostinger checkout page:

• Select the KVM 2 plan
• Choose a billing period (24 months is usually the lowest monthly price; 12 months is a good compromise if you want to keep upfront cost lower)
• Make sure “Hermes Agent auto deploy” is selected (this pre-installs Hermes for you)
• Pick the server location closest to you (for low latency)
• Optionally enable daily backups (weekly backups are on by default)

Complete your account registration, billing details, and payment. Hostinger includes a 30-day money-back guarantee if you run into issues early on.

Letting Hostinger auto-deploy Hermes

After checkout, Hostinger shows a short setup form for Hermes:

• It pre-fills an admin username (usually “Hermes”)
• It generates a strong password

Click the eye icon to reveal the password, then copy both the username and password and store them somewhere safe. These credentials control access to your Hermes web terminal, so don’t share them.

Click “Deploy” and give Hostinger a few minutes to build the VPS, set up Docker, and auto-install Hermes Agent with Traefik as a reverse proxy. When it’s done, you’ll land in Hostinger’s Docker Manager, where you’ll see a Docker project named something like hermes-agent-xxxx already running.

Opening the terminal and entering the Hermes container

Next you’ll use Hostinger’s built-in browser terminal to talk directly to the Hermes container.

1. In Docker Manager, copy the last four characters of the Hermes project name (the part after hermes-agent-).
2. Click the “Terminal” button in the top-right to open a new tab with a black terminal window. You’re already logged into the VPS.

Now, in the terminal:

1. Change into the Hermes project directory:
cd /docker/hermes-agent-XXXX
Replace XXXX with the four characters you copied.

2. Enter the Hermes container shell:
docker compose exec -it hermes-agent /bin/bash

Inside the container, you’ll run Hermes commands directly.

Running the Hermes setup wizard

Start the setup wizard with:

hermes setup

The wizard walks you through a quick configuration. When asked, choose the quick setup option.

Selecting your AI provider (OpenRouter)

Hermes needs an AI provider to power its “brain.” The guide uses OpenRouter because it gives you access to many models through a single API key and billing account.

In the wizard:

• Use the arrow keys to highlight “OpenRouter”
• Press Enter

The wizard will now ask for your OpenRouter API key. Leave the terminal open and switch to your browser.

Creating an OpenRouter account and API key

1. Go to OpenRouter.ai and sign up or sign in.
2. Verify your email if prompted.
3. Add credits to your account (the default $10 is plenty to start).
4. In the left menu, click “API Keys,” then “New Key.”

Give the key a name (for example, “Hermes Agent”). Optionally set a spending limit (e.g., $20 per month) as a guardrail so Hermes can’t exceed that amount with this key.

Click “Create.” You’ll see a long string starting with sk-. This is your API key. Copy it and store it somewhere safe—you won’t be able to see it again after closing the dialog.

Return to the terminal, paste the key into the wizard (you won’t see it appear for security reasons), then press Enter. The wizard confirms that the key is saved.

Choosing a default AI model (DeepSeek V4)

The wizard now asks which model Hermes should use by default. A model is the specific LLM that will handle reasoning and tool use.

For a beginner-friendly, low-cost setup, choose DeepSeek V4. It’s cheap per million tokens and works well with tool calls and multi-step workflows, making it a solid default for an agent. Use the arrow keys to highlight DeepSeek V4 and press Enter.

(If you want to go deeper into model choices for Hermes, check out this guide to top AI models for Hermes Agent.)

Keeping the local backend and enabling messaging

When the wizard asks where to run your terminal backend, accept the default “current local” and press Enter.

Next, it asks if you want to connect a messaging platform. Press Enter to proceed. You can connect Telegram, Discord, Slack, WhatsApp, and more, but Telegram is the fastest to set up and works well on all devices, so we’ll use that.

Connecting Hermes to Telegram

Hermes talks to Telegram through a Telegram bot. A bot is just an account controlled by software instead of a human.

Creating a Telegram bot with BotFather

1. Open Telegram on your phone or desktop and sign in.
2. Search for “BotFather” (the verified account with a blue check).
3. Start a chat and send /newbot (or tap the command).

BotFather will ask:

• A name for your bot (this is what you’ll see in chats)
• A unique username ending in “bot” (for example, MyHermesHelperBot)

Once accepted, BotFather replies with a message that includes a long token string. This is your bot token for the Telegram HTTP API.

Copy the token, return to the Hermes setup wizard in the terminal, paste it when prompted for the Telegram bot token, and press Enter.

Locking down who can talk to your agent

By default, anyone who finds your bot could message it and spend your AI credits. To prevent that, Hermes uses an allow list of Telegram user IDs.

To add yours:

1. In Telegram, search for “userinfobot” and start a chat.
2. It replies with your details, including your numeric ID.
3. Copy that ID and paste it into the Hermes wizard when asked which Telegram accounts are allowed to talk to your agent.

The wizard then asks if you want this ID to be your “home channel,” where Hermes sends scheduled job results and proactive messages. Type Y and press Enter.

The setup wizard now completes and shows a summary of your configuration, tools, and useful commands.

Launching the Hermes chat interface and gateway

To open the Hermes chat interface inside the terminal, run:

hermes

You’ll see information about the current model, tools, and installed skills at the top.

Next, you need to start the “gateway,” which listens for messages from Telegram and forwards them to Hermes.

1. Install the gateway integration:
hermes gateway install

2. Try to run the gateway:
hermes gateway run

If you get a permission error because you’re running as root, Hermes will suggest a safer command to run as the hermes user. Use:

sudo -u hermes /opt/hermes/.venv/bin/hermes gateway run

Once it starts successfully, the gateway is live and listening for Telegram messages.

Sending your first message from Telegram

Open Telegram and either:

• Search for the bot username you created, or
• Click the direct link to your bot from the BotFather message

Tap “Start.” You may see “Unknown command” in response to /start, which is normal—Hermes doesn’t use a built-in /start command.

Now send a real message, for example:

“You’re running on a Hostinger VPS inside Docker. Tell me what tools and skills you currently have available.”

Hermes will think for a moment and reply with a list of tools and skills. If you get a reply, your agent is fully wired up.

If you don’t, go back to the terminal and run:

hermes doctor

This checks for common issues like a bad API key, missing Telegram user ID, or gateway problems. If needed, you can re-run hermes setup to walk through the wizard again.

Using Hermes skills and web research

Two core features make Hermes more than a simple chatbot:

• Skills: saved workflows that Hermes can run again and again without rethinking every step
• Memory: persistent knowledge about you and your preferences that carries across conversations

To see this in action, you can ask Hermes to do real web research. For example, from Telegram:

“Find the top five YouTube videos ranking right now for ‘email marketing tutorial for beginners.’ For each one, save the title, channel, and what it covers to a file. Then tell me which gaps you noticed.”

Hermes may choose a built-in skill (for example, youtube-content), install a browser like Chrome if needed, and then:

• Open multiple pages on the web
• Extract structured information (titles, channels, topics)
• Analyze overlaps and gaps
• Save everything to a file on your server (e.g., /opt/data/email-marketing-top-five-youtube.md)
• Send you a summary in Telegram

The important part: the file lives on your VPS, under your control, and persists across sessions.

Turning workflows into reusable skills

Once Hermes has completed a complex task you like, you can turn that workflow into a named skill so you don’t have to retype the full prompt.

For example, after the YouTube research run, send:

“Save that workflow as a skill called competitor-watch. From now on, when I give you a search term, find the top five YouTube videos, save the same info to the file, and call out the gaps.”

Hermes will:

• Confirm it created the skill
• Describe the trigger (“when you give me a search term…”)
• Outline the steps it will follow
• Store any scripts it wrote (like a reusable Python script)

Skills you create from your own work are safe by default. If you ever install third-party skills, Hermes will scan them for unsafe code, but you should still only use skills from sources you trust.

If you want to explore more advanced Hermes setups, including multiple models and agent swarms, see this guide to running multiple AI models and agent swarms in Hermes Workspace.

Teaching Hermes about you with memory

Hermes can remember facts and preferences about you and use them in future tasks. This is especially powerful when combined with skills.

For example, you might say:

“Remember that I make YouTube tutorial videos about software, mostly for complete beginners. When you analyze gaps in competitor research, frame them in terms of what would help a complete beginner specifically.”

Hermes will confirm that it saved this as memory. To check what it knows, you can ask:

“What do you remember about me?”

It will list stored details, such as:

• Your name or role
• The type of content you create
• Your target audience (e.g., beginners)
• Context about the environment (Hostinger VPS, Telegram, etc.)

Over time, as you add more preferences and constraints, Hermes can tailor its research, writing, and automation to your style and goals without you repeating instructions every time.

Scheduling jobs so Hermes works while you sleep

One of Hermes Agent’s most useful features is scheduled jobs (cron jobs). These let Hermes run skills automatically at specific times and send you the results, even if your laptop is off.

You don’t need to write cron syntax. You just describe the schedule and task in plain language.

For example, to automate the competitor research skill:

“Every Monday at 9:00 a.m., run my competitor-watch skill on three search terms: ‘email marketing tutorial for beginners,’ ‘WordPress tutorial for beginners,’ and ‘SEO tutorial for beginners.’ Send me a summary on Telegram with the gaps that stood out.”

Hermes will:

• Parse the schedule (“Every Monday at 9:00 a.m.”)
• Confirm or ask for your time zone (e.g., “US Central Time”)
• Create a scheduled job that runs the skill with those search terms
• Send you a confirmation and the next run time

To see your scheduled jobs, just ask:

“Show me my scheduled jobs.”

Hermes will list each job with its description and schedule. You can also say “Run that job now” to trigger a scheduled job immediately and preview its output. If the format isn’t what you want, tell Hermes how to adjust it; it will refine the job for next time.

Over weeks, a job like this can build up a rich research file on your server while you focus on other work.

Managing AI and server costs

You have two main cost centers with this setup:

• OpenRouter usage (AI model calls)
• Hostinger VPS subscription

Tracking OpenRouter usage and switching models

In your OpenRouter dashboard:

• Go to “API Keys” to see usage per key
• Click “Activity” to see requests, tokens used, and spend over time (by hour, day, etc.)

Your choice of model is the biggest factor in AI cost. DeepSeek V4 is a good low-cost, capable default. If you need more reasoning power for a specific task, you can switch models from Telegram by sending the /model command. Choose OpenRouter, then pick a different model like Claude Opus for heavy reasoning, keeping in mind that more powerful models cost more.

You can also tell Hermes to use a specific model only for a particular scheduled job, so routine tasks stay cheap while complex ones get extra brainpower.

Checking your Hostinger billing

In the Hostinger dashboard:

• Click your profile in the top-right
• Go to “Billing” to see past invoices and upcoming renewals

Pay attention to the renewal price after your initial term, as introductory discounts may change. While you’re getting started, it’s smart to check your OpenRouter activity every few days so you can adjust model choices or limits before any surprise bills.

What to ask Hermes to do next

Once the basics are running, the easiest way to discover what Hermes can do is to ask it directly. For example:

“What kind of tasks are you good for?”

Hermes will typically suggest things like:

• Hands-on execution (writing, editing, coding, scripting)
• Long-running background work (research, monitoring)
• Scheduled recurring tasks (reports, summaries, checks)
• Code generation and refactoring
• Research and synthesis across multiple sources
• Memory-based workflows that adapt to your preferences

From there, you can start building your own skills and jobs: daily briefs, weekly reports, price monitoring, content research, or anything else you can describe in a sentence or two.

With Hermes Agent running on a Hostinger VPS, wired to OpenRouter and Telegram, you now have a personal AI agent that can learn how you work, remember what matters, and quietly get things done for you in the background.