Setting up OpenClaw on Hostinger takes about a minute with the managed one-click app, or about half an hour if you self-host it on a VPS with the Docker template. Both routes end at the same place: a working AI assistant you can talk to from WhatsApp or Telegram.
This guide walks through both setups using Hostinger’s official documentation, then covers the first-agent configuration — pairing messaging, choosing a model, and sending your first commands. Undecided between the two routes? Read Managed vs VPS first.
Before you start: what you need
- A Hostinger account (either route).
- For messaging: a phone with WhatsApp or Telegram installed.
- For the managed route: nothing else — AI credits are pre-integrated.
- For the VPS route: basic comfort with SSH, and either your own AI provider API key (OpenAI, Anthropic, Gemini, DeepSeek) or nexos.ai credits for model usage.
Route A: Managed OpenClaw (1-click, ~1 minute)
The managed route is the one Hostinger documents as “agent running within 60 seconds.” There is no server, no Docker, and no terminal.
- Get the plan. Order the AI automation app plan — — from Hostinger. Choose your term (24 months is the lowest monthly rate) and complete checkout.
- Open the app. In hPanel, find OpenClaw in the left sidebar, or go to the AI automation apps section and select OpenClaw. The system prepares your isolated container automatically.
- Choose your AI credits add-on. At onboarding you can enable “Ready-to-use AI” (nexos.ai credits for running language models) and “Instant web search” (Oxylabs credits for web search and data extraction). If you skip Ready-to-use AI, you will paste your own provider API key instead.
- Connect a messaging app. From the Hostinger dashboard — no separate OpenClaw dashboard needed — link WhatsApp by scanning the QR code (WhatsApp > Settings > Linked Devices) or connect Telegram with a BotFather token. You can also connect a Hostinger Email address for agentic email.
- Say hi. Click OpenClaw to open the interface, or copy the Access key token for CLI use. Send “Hi” in the chat: the agent starts an onboarding interview and walks you through the rest itself.
From here on, Hostinger handles security, updates, and backups. You manage the agent from the hPanel OpenClaw section, and you can switch to Hermes Agent, n8n, or Paperclip via tabs under the same subscription.
Route B: OpenClaw on a VPS (Docker template, ~30–60 minutes)
The VPS route gives you the same OpenClaw with full root access. The one-click template removes the Docker boilerplate, but you own everything after deployment.
- Order a VPS. KVM 1 (4 GB RAM) is enough for a standard single-agent setup; pick KVM 2 (8 GB) if you plan browser automation or multiple agents. 63% OFF Verified Oct 8, 2026
VPS KVM 2
$8.99/mo$24.49/mo24-month plan, paid upfront
Renews at $14.99/mo
- 2 vCPU + 8 GB RAM
- 100 GB NVMe storage
- 8 TB bandwidth
- Full root access
- n8n one-click template available
- Deploy the template. In hPanel go to VPS > Manage Server > Docker Manager > Catalog, search for OpenClaw, and click Deploy. Wait for the container to report healthy.
- Save the gateway token. The deployment shows an access token — copy it somewhere safe. It is your key to the OpenClaw gateway.
- Add your AI provider key. In the OpenClaw configuration, add your API key for your chosen model provider (Anthropic, OpenAI, Gemini, DeepSeek, or a local Ollama endpoint). Alternatively, fund nexos.ai credits for model usage without a provider account.
- Open the firewall. In hPanel > VPS > Firewall, make sure the ports OpenClaw needs are reachable, and point your domain’s DNS at the VPS IP if you are using one.
- Pair messaging. Configure the WhatsApp or Telegram bridge per OpenClaw’s documentation, then send your first message to confirm the agent responds.
- Set a restart policy. Make sure the container restarts automatically after a server reboot (
restart: unless-stoppedin compose, or the--restartflag), or your agent will silently die on the next maintenance window.
First-agent configuration: the essentials
Whichever route you took, the first hour with your agent looks the same. Work through these in order:
1. Pick your model deliberately
OpenClaw lets you bring any model. The trade-off is simple: frontier models (Claude, GPT) reason better and cost more per token; lighter models (Gemini Flash, DeepSeek) are dramatically cheaper and fine for routine tasks. On the managed plan, nexos.ai publishes per-model credit rates per million tokens, so you can compare before you commit. Start with a capable model while you learn, then downgrade the routine work once you see your usage pattern.
2. Connect one messaging channel well
Do not connect WhatsApp, Telegram, Slack, and Discord on day one. Pick the app you actually live in, pair it cleanly, and test a full round trip: ask the agent something, confirm it answers in the right chat, and check that it does not reply in the wrong one. Multi-channel comes later.
3. Give it something real to do
Agents configured in the abstract stay abstract. Give yours one concrete recurring job in the first session — a daily briefing, inbox triage, a website change monitor — and watch how it handles it for a week. That week teaches you more about prompts, skills, and heartbeat settings than any documentation.
4. Understand the heartbeat
OpenClaw can wake itself on a schedule (the “heartbeat,” every 30 minutes by default) to check for things to do. Each wake-up consumes model credits even when there is nothing to do. If your credit burn surprises you, the heartbeat interval is the first setting to lengthen, or switch the heartbeat to a lighter model.
5. Add skills one at a time
OpenClaw’s plugin system (“skills”) adds capabilities like browser control and file access. Install one, learn what it changes about the agent’s behavior, then add the next. A common beginner mistake is enabling everything at once and then not knowing which skill caused a strange behavior.
Common setup mistakes to avoid
- Skipping the API key / credits step on VPS. The template deploys fine without a model configured — and then the agent cannot think. This is the number-one “it deployed but does nothing” cause.
- Wrong WEBHOOK/public URL. If messaging webhooks never arrive, the agent’s configured public URL almost certainly does not match reality. Fix the URL, restart, re-pair.
- Firewall locked down. If you hardened the VPS firewall before deploying, messaging bridges and webhooks cannot reach the container. Open only what OpenClaw needs.
- No persistent volume (VPS). Confirm the data volume is mounted so your agent’s memory and config survive container recreations.
- Forgetting the gateway token. Save it at deploy time; recovering access later is annoying.
Useful next steps after setup
Once the agent answers its first message, resist the urge to automate your entire life in one evening. The setups that stick follow a pattern:
- Week one: one job. Give the agent a single recurring task — a morning briefing, a daily digest of a topic you follow, a reminder to review your calendar. Learn how it phrases questions, how reliable its timing is, and how it handles ambiguity.
- Week two: connect a second channel or skill. Add the file skill so it can read documents, or connect the email address so it can triage your inbox. One addition at a time, as with skills.
- Week three: scheduled tasks. Move beyond reactive chat into proactive work: heartbeat-driven checks, scheduled summaries, website monitors. By now you know your credit burn pattern, so the cost is predictable.
- Ongoing: review the credit balance monthly. A two-minute check in hPanel keeps billing boring, which is exactly what you want from infrastructure.
If something breaks along the way, the troubleshooting guide covers the common failures in the order you are most likely to hit them.
FAQ
How long does it take to set up OpenClaw on Hostinger?
The managed 1-click app takes about a minute from hPanel with no technical steps. The VPS Docker template takes roughly 30–60 minutes for a first-timer including API key configuration and messaging pairing.
Do I need my own OpenAI API key for managed OpenClaw?
No. The managed plan comes with pre-integrated nexos.ai credits for models (and Oxylabs credits for web search), topped up from the dashboard. You can optionally connect your own API key or a ChatGPT Plus/Pro subscription instead.
Which VPS plan is enough for OpenClaw?
KVM 1 (1 vCPU, 4 GB RAM) handles a standard single agent. Choose KVM 2 (2 vCPU, 8 GB RAM) if you will use browser automation skills or run multiple agents on the same server.
Can I use WhatsApp with OpenClaw?
Yes, on both routes. Managed: scan the QR code from the Hostinger dashboard (WhatsApp > Settings > Linked Devices). VPS: configure the WhatsApp bridge per OpenClaw’s docs and pair the same way.
What is the OpenClaw heartbeat and why does it cost credits?
The heartbeat wakes the agent on a schedule (default every 30 minutes) to check for tasks. Each wake-up runs the model, consuming credits even when idle. Lengthen the interval or use a lighter model to cut the cost.
Can I switch from OpenClaw to Hermes Agent later?
On the managed plan, yes — both apps (plus n8n and Paperclip) are included in the one subscription and you switch via tabs in hPanel. Note that switching starts the new app fresh; configurations do not transfer.