Before you start
You do not need to be a developer for either route, but they assume different things. The managed route assumes nothing — Hostinger says “no technical setup,” and the onboarding is visual and guided. The self-hosted route assumes you are comfortable with a KVM VPS, a Docker Manager panel, and pasting API keys; root access is yours, and so is the responsibility.
- Managed route: a Hostinger account and the AI automation plan. That is the whole list.
- VPS route: a Hostinger KVM VPS — Hostinger’s own tutorial recommends at least 2 vCPU and 8 GB RAM for a stable Docker deployment. An admin username and password of your choosing, plus (optionally) API keys for your model provider or Hostinger’s AI Router credits.
One honest note up front: Hermes is open-source software by Nous Research, not by Hostinger. Hostinger sells hosting and the managed experience around it. If you run into a bug in the agent itself, that is upstream software — Hostinger’s job is the infrastructure, updates to the template, and support for their platform.
Managed setup: step by step
Per Hostinger’s managed product page (read October 9, 2026), the managed setup is designed to take about 60 seconds. Here is the flow:
- Buy the AI automation plan through the managed apps page and choose Hermes as your starting app. The plan covers four apps (Hermes, OpenClaw, n8n, Paperclip) — you pick one to begin.
- Choose your term (24 months is the cheapest per-month rate; 12-month and month-to-month terms are available at higher rates) and decide on add-ons: ready-to-use AI credits (nexos.ai) and web search credits (Oxylabs) can be added at checkout.
- If you skipped ready-to-use AI credits, paste your own model provider API key during onboarding (OpenAI, Anthropic, and similar are supported).
- Connect a messaging channel. Telegram and WhatsApp are the first-class options — the guided onboarding walks you through pairing them from your Hostinger dashboard. No separate Hermes dashboard is needed.
- Choose your AI model through the Hostinger AI Router or bring your own key. The router gives you access to Claude, Gemini, and GPT-class models without separate accounts.
- Finish the guided onboarding and say “Hi” to your agent. Hostinger prepares the environment behind the scenes; you get a visual interface plus CLI access.
That is it. From here, long-term memory, web search, and managed security, updates, and backups are all handled by Hostinger. You never touch a server.
VPS setup: step by step
Prefer full control? Per Hostinger’s official Hermes-on-VPS setup guide, the one-click template flow is:
- Order a KVM VPS — again, Hostinger’s tutorial recommends 2 vCPU / 8 GB RAM minimum for a stable Docker deployment.
- In your VPS dashboard, go to Docker Manager → Catalog, search “Hermes Agent,” and hit Select.
- Set an admin username and password. Write them down somewhere safe — this is the admin gate to your agent.
- Optionally connect Hostinger AI Router credits (auto-filled if you bought them with the VPS) and/or Oxylabs credits for web search and scraping. You can also enter your own API keys manually at this point.
- Click Deploy. Hostinger creates and starts the Hermes container plus supporting services like Traefik.
- Open it with the “Open” button next to the Hermes project — this launches the Hermes CLI.
- Work through the first-run CLI setup (arrow keys + Enter/Space): choose Full setup (your own providers, including Hostinger AI/Oxylabs credits) or Quick Setup (Nous Portal sign-in, no API keys); select your model provider; toggle communication channels (Telegram, WhatsApp, others); and enable tools — check Web Search & Scraping and choose Oxylabs as provider if you bought credits.
Type /help in the CLI to see the available commands. And remember: on this route, updates, security hardening, and recovery are yours to manage — the template gets you deployed, not maintained.
First tasks to give your agent
Hermes is built for task execution and autonomous workflows, not just chat. Once messaging is connected, try these as shakedown tasks — each one exercises a different capability:
- A research brief: “Find five recent articles about [your topic], summarize each in two sentences, and save the summaries to a file.” This tests web search, reading, summarization, and file writing in one go.
- A code task: “Write a Python script that reads this CSV and emails me a weekly summary.” This tests code generation and execution, which is where Hermes differs most from a plain chatbot.
- A recurring monitor: “Check this price page every morning and message me on WhatsApp if it drops below X.” This tests scheduling, persistence, and the messaging connection.
- A learning loop: “Draft my reply to this support thread, then improve your draft based on my edits.” Hermes is described as self-improving — corrections you make feed back into how it handles similar tasks.
Start small and cheap. Run your first tasks on a light model like Gemini Flash so you learn the agent’s behavior before spending heavier-model credits on production work. And watch your credit balance for the first week — autonomous agents can burn through credits faster than expected if a task loops.
Connecting Telegram and WhatsApp
Messaging is how most people will actually use Hermes day to day, so it is worth getting right. On the managed plan, pairing happens from your Hostinger dashboard during onboarding — no bot-building required on your side. On the VPS route, the first-run CLI setup includes a channel toggle step where you enable Telegram, WhatsApp, and any other supported channels.
Security note: your agent can act in the world — sending messages, running code, writing files. Treat the messaging connection like a credential. Do not add the agent to large group chats where strangers could prompt it, and keep the admin password you set at deploy time out of chat logs.
How updates work
This is where the two routes genuinely diverge, and it is worth understanding before you commit.
- Managed: Hostinger handles updates, security patches, and backups. The managed container runs an isolated instance of the agent, and Hostinger’s team rolls out updates to the platform. You do not schedule or apply anything — the trade-off is that you do not control the timing either.
- VPS: the Docker template gets you a working deploy, but the container is yours. You pull updates, patch the OS, manage backups, and recover from failures. If the Hermes project ships a new version with a feature you want, you upgrade it — on your schedule, with your rollback plan.
If you outgrow the managed plan but do not want to start over, note that Hostinger offers one free migration from Managed OpenClaw to a VPS (KVM 1) within 30 days of purchase, via a Hostinger agent — but there is no equivalent migration path between Managed Hermes and Managed OpenClaw. For Hermes specifically, a VPS move means a fresh deploy of the template, not a settings transfer.
Common setup problems
- Agent stops responding: the most common cause is an empty AI-credit balance. Credit balance can never go below zero — requests are declined when it runs low. Top up from the Hostinger dashboard (managed) or Docker Manager → Projects (VPS), then retry.
- Messaging pairing fails: re-run the pairing flow from the dashboard rather than retrying in chat. On Telegram, a fresh bot token fixes most pairing issues; on WhatsApp, unlink and re-link the device.
- Web search returns nothing: web search runs on Oxylabs credits, which are separate from model credits. If search was never enabled at setup (the Web Search & Scraping tool checkbox), enable it and make sure Oxylabs credits are present.
- VPS deploy hangs: the usual culprits are an undersized VPS (below Hostinger’s 2 vCPU / 8 GB recommendation) or a Docker Manager hiccup. Give it a few minutes, then check the Docker Manager project status before redeploying.
FAQ
On the managed plan, Hostinger says about 60 seconds — pick the app, connect messaging and a model, and the guided onboarding finishes the job. The VPS route takes longer: ordering the VPS, deploying the Docker template, and working through the first-run CLI setup typically takes 15–30 minutes if nothing goes wrong.
Not on the managed plan — the onboarding is visual and guided, with no terminal required. On the VPS route you will interact with the Hermes CLI and may need to paste API keys, but the template handles the Docker plumbing. Full root access means you should be comfortable troubleshooting a Linux container when something breaks.
Full setup lets you configure your own providers — your model provider (including Hostinger AI Router credits) and Oxylabs credits for web search. Quick Setup signs you in through the Nous Portal instead, which needs no API keys of your own. Full setup gives you more control; Quick Setup gets you running faster.
Yes. The setup flow includes a channel toggle step where you can enable multiple communication channels. On the managed plan, pairing is done from the Hostinger dashboard during onboarding; on the VPS route, channels are toggled during the first-run CLI setup.
On the managed plan, yes — Hostinger handles updates, security patches, and backups as part of the service. On the self-hosted VPS route, no — you own the container, so you pull updates, patch the OS, and manage backups yourself.
Check your AI-credit balance first: when credits run out, the agent stops generating responses until you top up. The balance can never go below zero, so a silent agent almost always means an empty balance. Top up from the Hostinger dashboard (managed) or Docker Manager → Projects (VPS).
Still deciding between routes? Our Hermes Agent guide covers pricing and the honest trade-offs, Hermes pricing breaks down the true monthly cost, and AI agents on Hostinger compares Hermes with OpenClaw.