Become a TechRadar Insider
Running an AI agent on a third-party platform means your data, API keys, and conversation history sit on someone else's server. For developers and small teams who want real privacy or tighter cost control, that's a legitimate concern. Self-hosting OpenClaw on a VPS is a practical fix, and the process is more straightforward than it used to be.
OpenClaw is an open-source AI assistant that connects to messaging apps like WhatsApp, Telegram, and Slack, and lets an AI model handle real tasks on your behalf. Because the gateway runs on your own server, you decide which AI providers to connect, which Skills to enable, and how much system access the agent gets.
What you need before you start
You'll need an API key from your chosen AI provider before provisioning a server. OpenClaw works with Anthropic's Claude, OpenAI's GPT models, and local models via Ollama, among others. Telegram is the easiest messaging platform to start with on a VPS, since it doesn't require port forwarding.
On the hardware side, OpenClaw's documentation recommends at least 2 vCPUs and 4 GB RAM for development and testing. Production workloads are better served by 4 vCPUs and 8 GB RAM, with at least 20 GB of SSD storage. If you plan to run local AI models via Ollama alongside OpenClaw, budget for significantly more RAM.
How to self-host OpenClaw on Hostinger VPS
Hostinger is our recommended VPS pick for OpenClaw in 2026. It's one of the few providers with a dedicated one-click OpenClaw template built into its control panel, which removes a significant chunk of manual work around Nginx, SSL certificates, and Docker setup.
Step 1: Choose your plan.
Head to Hostinger's OpenClaw VPS hosting page. Our recommended KVM2 plan starts at around $8.99/month on a two-year term. The entry-level configuration covers most personal or small-team workloads, but if you expect heavier automation or multiple concurrent users, move up a tier.
Step 2: Add your AI credits or API key.
During checkout, you can add nexos.ai credits directly to your plan, giving you access to OpenAI and Anthropic models without managing API keys separately. Bringing your own API key is also supported at this stage.
Step 3: Save your gateway token.
Once your order is confirmed, copy your OpenClaw gateway token and store it somewhere secure. You'll also connect your WhatsApp number or preferred messaging platform at this step.
Step 4: Deploy.
Hostinger provisions your VPS, installs OpenClaw via Docker, and configures the environment automatically. You'll land in the Docker Manager inside hPanel when it's done.
Step 5: Configure your instance.
Access your OpenClaw instance through the Control UI or an SSH tunnel from the Docker Manager. The first account you create becomes the administrator, and you can configure Skills, adjust active messaging channels, and tune performance settings from there.
Shared hosting can't support OpenClaw's always-on architecture, so the dedicated CPU and memory allocation on a VPS matters more than the spec sheet suggests. Hostinger's network is also well-suited to the persistent WebSocket connections OpenClaw uses across messaging channels.
Securing your deployment
Security is your responsibility when you self-host, so it's worth sorting a few things before you go live. Keep OpenClaw's gateway port (18789) off the public internet and route external traffic through Nginx instead. The gateway should bind to 127.0.0.1, not 0.0.0.0, which would leave it exposed without authentication.
Access the Control UI remotely via an SSH tunnel or Tailscale rather than binding it to a public IP. OpenClaw has reach over your files, shell, and browser, so a misconfigured instance is a real attack surface. If you're running it for a team where not everyone shares the same trust level, separate gateway instances are safer than a shared one.
We also recommend setting up automated backups of your workspace and configuration files before anything else. Hostinger includes backup options across its VPS plans, but check the schedule and test a restoration before you actually need it.
Other providers worth considering
Hostinger isn't the only provider with OpenClaw support. A few others are worth looking at, depending on your budget and location:
- Contabo has dedicated OpenClaw VPS plans with defined resource tiers starting from 4 vCPUs and 8 GB RAM, with competitive pricing on higher-spec configurations.
- Kamatera bills hourly and offers 20+ global data centers, which suit teams that need granular cost control or low latency to a specific region. Development environments start at around $6/month.
- DigitalOcean is a solid choice for manual deployments. The official OpenClaw docs cover DigitalOcean setups, and the droplet marketplace has community-built configurations to start from.
- Hetzner is popular in the OpenClaw community, especially for European workloads, and is one of three providers supported by ClawHost, an open-source tool for automated OpenClaw deployments.
- Vultr works well for self-managed setups and is also supported by ClawHost alongside Hetzner and DigitalOcean.
Across all of these, the baseline requirements are the same: KVM virtualization (shared hosting won't run Docker), 4 vCPUs and 8 GB RAM for production, and full root access.
What to expect once you're live
Once your instance is running, OpenClaw stays online around the clock and responds to messages or executes tasks whether or not your own machine is on. Individual users typically start with scheduling, email triage, and web research through simple chat commands. Teams tend to build more involved workflows over time, like internal bots, GitHub automation, and multi-channel support tools.
OpenClaw is still a fast-moving project, and what it can actually do depends on which Skills and AI providers you connect. Before enabling any third-party Skills on a production server, review their code. Check periodically which file system permissions and tools your agent needs, and remove anything it doesn't use. Broader access than necessary is a liability.
You must confirm your public display name before commenting
Please logout and then login again, you will then be prompted to enter your display name.