Your own AI agent,on your own server.
Pick a model and a channel. We rent the server, install OpenClaw and bring it online.
Private beta · no card to join
Self-hosting an agent is an afternoon of ops.
By hand
- Rent a VPS, add SSH keys
- Install Node.js and Ollama
- Pull a model, wire API keys
- Install OpenClaw, set the bot token
- Keep it alive with PM2
- Debug why it stopped replying
With HyperClaw
4 steps
Model, channel, token, deploy.
The wizard
Four steps. No terminal.
During the private beta we run each deploy together with you.
Pick a model
Claude, GPT-4, Gemini, or a local Llama / Qwen / Phi-3.
Pick a channel
WhatsApp, Telegram or Discord.
Add your token
Bot token, server provider, optional docs.
Deploy
Watch it provision until the agent replies.
One server. One tenant. You.
Pick a local model and your chats never touch a third-party model API.
Read more
With a local model, HyperClaw provisions a VPS for you alone. There is no shared inference endpoint. The model runs under Ollama on that machine, and OpenClaw is kept alive by PM2. Cloud models (Claude, GPT-4, Gemini) are also available when you want more capability than a small local model gives.
What you get
Cloud or local model
Cloud · API
Local · your VPS
Get a setup slot.
Private beta. We email you when a slot opens and set it up with you.