Agent Zero, OpenClaw, n8n and Flowise: the self-hosted AI agents you can deploy in one click
Four very different things get called an AI agent. What each one actually does, how much RAM it needs, and the security question that the one-click install pages do not ask you.
Four things are commonly called self-hosted AI agents, and they are not variations on one idea. They are genuinely different tools with different risks. Picking the wrong one wastes an afternoon; deploying one of them carelessly can cost you the server.
n8n — a workflow tool that can call a model
The oldest idea of the four and the safest. n8n is a visual automation tool: you draw a flow of triggers and actions, and one of the available actions is calling a language model. An agent built in n8n is a workflow with a model in the middle.
- Best for: connecting things that already exist — a form, a spreadsheet, an inbox, a CRM — with a model somewhere in the chain.
- Runs on: 1 GB RAM and 1 CPU minimum, but Hostinger's own guide recommends 2 GB and 2 cores for headroom.
- Listens on: port 5678.
- Risk: low. It does what your flow says and nothing else.
Flowise — the same idea, built for chains
Also visual, also self-hosted, but built specifically around model chains and retrieval rather than general automation. If what you want is a chatbot over your own documents, Flowise gets you there with less scaffolding than n8n. If what you want is to move data between six services, n8n is the better fit.
OpenClaw — an assistant you talk to
A conversational agent you run yourself. Closer to a chat assistant with tools than to an automation platform, and the one most people mean when they say they want to self-host an AI assistant.
Agent Zero — a model with a computer
The most capable and the most dangerous of the four, and the difference is worth stating precisely. Agent Zero's own description is a framework that gives a language model access to a full computer environment. In practice the one-click deployment includes a search engine, an SSH server for code execution, and persistent memory.
The published capability list is the point: agents can execute terminal commands, write and run code, browse the web, manage files, and create new tools for themselves. That is not a chatbot. That is a model with a shell.
The question the install pages do not ask
The same caution applies in weaker form to the others. An n8n instance on an open port is a set of your credentials waiting to be used by whoever finds it, because the whole value of the tool is that it holds tokens for your other services.
- Give the agent its own VPS, not the one your site runs on.
- Bind to 127.0.0.1 and tunnel, or require authentication at a reverse proxy.
- Use a scoped API key for the model, not your account-wide one.
- Assume anything the agent can read, it can leak; anything it can run, it can be talked into running.
Where to run them
All four deploy in one click from Hostinger's VPS template library, which is the least painful route we have found — the template installs the Docker stack and you manage it from the panel. Their KVM 2 tier, at 2 vCPU and 8 GB RAM, is the one their own pages mark as most popular for these templates and is a sensible starting point for anything except a bare n8n flow.
One thing to know before you pick a plan: these are CPU-only machines. There is no GPU, which matters enormously if you were planning to run the model locally too. We go into that in the step-by-step guide.
Hostinger
20% off your first Hostinger plan
20% OFF
Sources
Capability and requirement descriptions from each project's own one-click template page on Hostinger, and from Hostinger's own n8n installation tutorial, read on 13 September 2026. The security guidance is ours, not theirs.
Affiliate Disclosure: Some links on KuponGuru are affiliate links. We may earn a commission at no additional cost to you.
More guides
how-to-guides · 7 min read
How to set up an AI agent on a Hostinger VPS, step by step
The one-click templates, the plan you actually need, the warning screen that deletes everything on your server, and the architectural mistake that makes people think a VPS is too slow for AI. A complete walkthrough.
Read article
ai-tools · 3 min read
Sora 2, Veo 3.1 and Kling: where you can actually use them, and what it costs
The three video models everyone is searching for, what access to each really requires, and why the honest answer to “how do I use Sora 2 free” is not the one the round-up pages give.
Read article
savings-guides · 3 min read
One AI subscription instead of six: what the all-in-one workspaces actually include
Platforms that resell every frontier model for one monthly fee are having a moment. The maths is real, the savings claim is inflated, and the cheapest tier almost never includes the model you came for.
Read article