Characters are original Nxeon illustrations, not the logos or mascots of the projects named.

AI agent hosting

Where AI agents live.

OpenClaw, Grok-powered bots, Claude Code, MCP servers and n8n, on a server that never sleeps. Each one gets its own virtual machine with full root, NVMe storage and a managed firewall. Plans from $8.88/mo for your first term, then $12.78/mo.

Deploy a server Watch them work

Characters are original Nxeon illustrations, not the logos or mascots of the projects named.

01The night shift

Six agents. One server. All night.

Agents work in parallel on one always-on VM while your laptop sleeps. This is an illustrated sample of what that looks like.

1Queue
2Dispatch
3Work
4Done

PinchReply to a chat

VexScrape and summarise

QuillDeploy a site

SocketAnswer a webhook

NodiRun a nightly job

BitTriage an inbox

0tasks done in this sample

Illustrative sample. The tasks, times and counter are made up to show agents working in parallel. They are not live data or a measure of real throughput.

02Meet the crew

Pick an agent. Watch it move in.

Choose a character to see a sample setup on a Nxeon server: the commands, the status lights, and a first reply.

OpenClawGrokClaude CodeMCPn8nyour codeIllustrative sample

Pinch

A personal agent that lives in your chat apps and does things for you.

root@your-server

$ curl -fsSL https://openclaw.ai/install.sh | bash
✓ installed (Node 24)
$ openclaw doctor
✓ gateway bound to 127.0.0.1
✓ telegram paired

03How an agent moves in

An idea.You want an agent that answers while you're away: your chats, your inbox, your workflows.
A server that never sleeps.A Nxeon VPS stays on with a fixed IP, NVMe storage and full root. It doesn't stop when your laptop lid closes.
The agent moves in.SSH in, run the project's own installer, then put it under systemd so it restarts by itself.
It runs while you sleep.It replies, scrapes, builds and sends overnight on its own VM, while your devices are switched off.
It keeps itself safe.Its own VM and a managed firewall put a shell around it, and snapshots let you roll back a bad change.

04What runs here

The rack. Pick a slot.

What works today and what's coming, each one labelled plainly. Sizes are our suggestion for one person's setup. Check each project's current docs, because requirements change.

OpenClaw

The open-source personal agent for Telegram, Discord, WhatsApp, Slack and more. Its docs say to keep the gateway on loopback and reach it over an SSH tunnel.

Self-install today
One-click template: coming soon

SuggestedStandard · 2 vCPU · 4 GB RAM · $12.99/mo to start

Deploy Standard Setup guide

Grok bots

A small Node or Python process for Telegram, Discord or X that calls xAI's API. The API works with the OpenAI SDK, and in long-poll mode the bot needs no open inbound port.

Self-install today
Starter template: coming soon

SuggestedStarter · 1 vCPU · 2 GB RAM · $8.88/mo to start

Deploy Starter Bot hosting guide

Claude Code

Already on every new VPS at first boot. Use it over SSH, or run claude -p from a systemd timer for unattended jobs. Bring your own Claude subscription or Anthropic API key.

Pre-installed on every new VPS

SuggestedStarter · 1 vCPU · 2 GB RAM · $8.88/mo to start

Deploy Starter How it works

MCP servers

Serve your own tools over Streamable HTTP behind HTTPS on your domain. The MCP spec asks remote servers to check the Origin header and to authenticate every connection.

Self-install today
MCP host template: coming soon

SuggestedStarter · 1 vCPU · 2 GB RAM · $8.88/mo to start

Deploy Starter Docker hosting

n8n

Webhook triggers need a public HTTPS address, and your workflows and credentials need a persistent volume. A VPS with your own domain gives it both.

Self-install today
Agent template: coming soon

SuggestedStandard · 2 vCPU · 4 GB RAM · $12.99/mo to start

Deploy Standard n8n hosting

Your own agent

Hermes Agent, LangGraph, a browser agent, anything. Full root means any runtime: add Docker when you deploy and bring your compose file. Browser automation needs room.

Docker add-on at checkout

SuggestedPro · 3 vCPU · 8 GB RAM · $20.99/mo to start

Deploy Pro Docker hosting

05Agents as customers

An agent asks. A server appears.

We're building a control plane an agent can drive safely: a scoped token, a budget it can't exceed, and a server on request. This is how it will work.

Coming soonPlanned API, shown as a preview. It isn't available yet.

Requestcreate_server {"plan":"standard"}
Scope checkthe token may create servers
Budget check$12.99 of a $30.00 example monthly cap
Provisionwallet balance only, never your card
Onlinefirewall rule added, SSH key in place

Available today

API keys in Settings, and GET /api/v1/servers to list your servers
A machine-readable summary of plans and live prices at /llms.txt
SSH keys you save are added to new servers at first boot

Coming soon

Scoped tokens and a full REST API with an OpenAPI spec
A Nxeon MCP server and a CLI with the same actions
Wallet budgets per token with a hard monthly cap

Developer overview

06Real VPS pricing

No AI-hosting markup.

Agents run on the same VPS plans as everything else. The model you connect is billed separately by the company that runs it.

Starter

$8.88~~$12.78~~first term

Then $12.78 a month

vCPU: 1
RAM: 2 GB
NVMe: 25 GB

Chat bots on a hosted model, Claude Code jobs, small MCP servers

Deploy Starter

Good starting sizeStandard

$12.99~~$18.70~~first term

Then $18.70 a month

vCPU: 2
RAM: 4 GB
NVMe: 50 GB

OpenClaw with a few channels, n8n with Postgres

Deploy Standard

Pro

$20.99~~$30.22~~first term

Then $30.22 a month

vCPU: 3
RAM: 8 GB
NVMe: 80 GB

Browser automation, several agents on one box

Deploy Pro

07FAQ

Straight answers.

Is there a one-click OpenClaw image?

Not yet. Today you deploy a clean Ubuntu server and run OpenClaw's official installer, which takes a few minutes. A hardened one-click template is coming soon, and we'll only say it exists once it does.

Does the server include the AI model or API credits?

No. Nxeon provides the server. Model usage, whether that's xAI, Anthropic, OpenAI or another company, is billed by that company under your own account. Running a large model locally needs far more memory than these plans have.

How much RAM does an agent need?

A chat bot that calls a hosted model is light: Starter (2 GB) is plenty. OpenClaw or n8n are more comfortable on Standard (4 GB). Browser automation uses a lot of memory, so start at Pro (8 GB). You can resize later.

Is it safe to run an agent that executes commands?

Run it in its own VM, not on your laptop or next to production. Keep its control panel off the public internet, open only the ports you need, treat third-party skills and plugins as untrusted code, and snapshot before you add one.

Can my agent create and manage its own server?

Partly. Today an API key can list your servers, and /llms.txt describes our plans. Creating servers, scoped tokens, an MCP server and per-token budgets are planned and marked "coming soon" above.

Is the "night shift" animation real data?

No. The tasks, clock, activity feed and counter are an illustrated sample showing how agents can run in parallel on one server. They aren't measurements of anything.

Are you affiliated with OpenClaw, xAI, Anthropic or n8n?

No. We name these projects only to describe software you can run on your own server. The names are trademarks of their owners, and the characters on this page are our own drawings, not their logos or mascots.

Your server is 47 seconds away.

Give your agent a home that stays on. Full root, your own VM, and real prices with no AI markup.

Deploy a server See VPS plans

OpenClaw, Grok, xAI, Claude, Anthropic and n8n are trademarks of their respective owners. Nxeon is not affiliated with or endorsed by them.