Your AI agent, working 24/7 from $4 a month
Spin up n8n or OpenClaw in 1 click on a server that never sleeps, from ~$4/mo, and connect it to any model: GPT, Claude, Grok or DeepSeek through CubePath's AI Gateway. One API, one bill.
Your apps and workflows live on your server · No vendor lock-in
One API for every model. One bill.
Your apps need a model. Instead of juggling a key per provider, connect them to CubePath's AI Gateway: GPT-4.1, Claude, Grok and DeepSeek behind a single endpoint, at the same price as going direct and with all your spend on one bill.
One endpoint, every model
OpenAI, Anthropic, xAI and DeepSeek with a single key. OpenAI-compatible SDK: point the base URL and you're done.
Unified cost
All your AI spend on one bill, at the same price as the provider. No platform fees, no markup.
Automatic failover
If a provider goes down, the Gateway routes to another. Your apps keep responding without noticing.
One dashboard for it all
Requests, costs and performance across every model in a single dashboard.
You pay the same as going direct to each provider, with no fees, but with one key, one bill and failover between models built in.
The AI apps you deploy in 1 click
Production-ready. Click, and they're running on your own server in minutes.
OpenClaw
Your AI agent on WhatsApp, Telegram or Discord, always on and running the model you pick via the AI Gateway.
Deploy Now
Hermes Agent
Autonomous self-hosted AI assistant with persistent memory — runs real tasks, builds its own skills and connects to Telegram, Discord, Slack, WhatsApp and email.
Deploy NowAI that works on its own, day and night
Apps and agents that run by themselves on your server, with the model wired to the AI Gateway.
24/7 assistant in your channels
With OpenClaw, an agent that replies on WhatsApp, Telegram or Discord at any hour.
Automatic tickets and replies
n8n classifies inbound and drafts replies by firing the model through the Gateway.
Summaries and alerts
A cron reads, summarizes with AI and pings you on Slack or email. Hands off.
Agent over your documents
Answer over your contracts, wikis and internal data with an n8n + RAG workflow.
Enrich and classify data
Tag, extract and normalize data in batches: from the CRM to the spreadsheet, automatically.
AI-powered webhooks
Receive an event and respond with a model instantly: from form to action.
Automated back office
Invoices, emails and CRM processed by AI while you do something else.
Processes that never sleep
Bots and workflows running 24/7 on your server, with no dependence on your laptop.
The server is yours. You're in charge.
Your apps, your workflows and your credentials live on your VPS, and you pick (and swap) the model whenever you want.
Your server, your data
Apps, workflows, credentials and databases live on your VPS, not in a third-party SaaS you don't control.
No vendor lock-in
Switch between GPT, Claude, Grok or DeepSeek without touching the code, and change servers whenever you want.
You choose where it runs
You decide the datacenter where the server lives, which keeps data residency and GDPR simple.
An always-on server for very little
Since the model runs on the AI Gateway, your server just orchestrates the apps, and that's lightweight. Start for almost nothing and size up as you scale.
Get started
One agent or a light workflow
Best for most
An agent plus everyday workflows
Your team
Several apps and more workflows
Production
Real traffic and concurrency
Scale
Multi-user and high volume
We recommend gp.micro to start: it runs n8n and a single agent comfortably. gp.nano is fine for testing or one light workflow; move up to gp.starter for OpenClaw at scale or several agents at once.
What devs ask before they set it up
Straight answers, no fluff.
How do I run an AI agent 24/7?
Deploy an AI app (an OpenClaw agent or n8n automations) on an always-on VPS. Unlike a script on your laptop, it keeps running 24/7 after you log off: replying to messages, processing events and running tasks on its own. You deploy it in 1 click and wire it to the AI Gateway.
What is an AI Gateway and why do I need one?
It's a single API that connects your apps to every model. Instead of one key per provider, you call one endpoint for GPT-4.1, Claude, Grok or DeepSeek, with automatic failover and all your spend on one bill. One key, one SDK, every model.
Is the AI Gateway more expensive than going direct to OpenAI or Anthropic?
No. You pay the same list price as going direct to each provider: zero markup, zero platform fees. What you gain is one key, one bill, a shared dashboard and automatic failover between models.
Which AI models can I use: GPT, Claude, Grok, DeepSeek?
All the majors: GPT-4.1 and GPT-4.1 mini (OpenAI), Claude Sonnet 4.5 and Haiku (Anthropic), Grok 3 and Grok 3 mini (xAI), and DeepSeek V3 and R1 (DeepSeek). You switch between them without touching your code.
Is the AI Gateway OpenAI-compatible? Can I reuse my code?
Yes. The AI Gateway is OpenAI-compatible: change the base URL and key, pick a model, and your existing chat/completions code keeps working, now against the Gateway instead of a single provider.
How do I put an AI agent on WhatsApp, Telegram or Discord?
Deploy OpenClaw in 1 click and connect it to the AI Gateway. It's an AI agent that lives on WhatsApp, Telegram or Discord and answers your users or your team 24/7 with whichever model you pick, all on your own server.
Can I self-host n8n on a VPS? Is it free?
Yes. n8n is free and open-source, and runs fully self-hosted: deploy it in 1 click from the Marketplace, live in minutes with no Docker setup. Your workflows and credentials stay on your server, with no per-execution limits.
n8n self-hosted vs n8n Cloud: which is cheaper?
Self-hosting wins on cost and control once you run more than a few thousand executions a month: no per-execution limits and no rising SaaS tiers, just a flat VPS fee from ~$8/mo. Cloud is faster to start; self-hosted keeps your data and credentials on your own server.
How much does it cost to run an AI agent 24/7?
From ~$4/mo with gp.nano for testing; we recommend gp.micro (~$8/mo) to run an agent or your workflows comfortably. VPS are billed hourly, so you only pay for what you use and can power off anytime; see the pricing page. Model usage is billed separately through the Gateway, at provider list price.
Do I need a GPU to run an AI agent?
No. The model runs on the AI Gateway, not on your server, so your VPS just orchestrates the apps, a light CPU plan is plenty. No GPU and no big RAM needed. You'd only need a GPU to run a model locally, which is a different setup.
Where does my data live? Is it GDPR-friendly?
Your workflows, credentials and databases live on your VPS, in the datacenter you choose, which keeps data residency and GDPR simple. Requests to the model do go to the provider through the Gateway, as with any AI API, but orchestration and your data stay under your control. For inference that never leaves your server you'd run a local model.
How do I connect n8n or OpenClaw to the AI Gateway?
Point the app at the AI Gateway endpoint with your key. Since it's OpenAI-compatible, in n8n use the OpenAI node (or an HTTP Request node) with the Gateway's base URL; in OpenClaw, configure it the same way. Then pick a model and fire inference from any trigger.
Put your AI to work in 5 minutes
Deploy OpenClaw or n8n in 1 click, connect it to the AI Gateway and leave your AI running day and night. Server from ~$8/mo and one bill for every model.