Self-host private Ollama inference and Open WebUI with no token fees, document RAG, AMD AVX-512 speed and an OpenAI-compatible API.
Billed yearly at $179.94. $29.99/mo when billed monthly.
Starting plans
This plan covers most first deployments. Move up only when sustained load shows you need more.
Included: SRV-B2C · 2 vCPU · 4 GB · 60 GB NVMe
You can upgrade in place later; your data and IP address stay the same.
Simple monthly prices. Pay yearly and save 50%.
The easy way to get started.
More room for growing projects.
Extra power for busy workloads.
Maximum resources for heavy use.
Deploy self-hosted Ollama local inference and Open WebUI on dedicated AMD EPYC NVMe cloud servers. Run DeepSeek-R1, Llama 3.3, Mistral, and Qwen 2.5 with zero third-party telemetry, document RAG, OpenAI API compatibility, and unlimited prompt tokens.
Self-host your private ChatGPT alternative and LLM inference engine without per-token subscription fees or third-party data tracking. Run modern quantized models (DeepSeek-R1, Llama 3.3, Qwen 2.5) with local nomic embeddings, multi-user role management, web search grounding, and OpenAI-compatible API routes.
Why a server that never sleeps works better.
| NodexVPS server | Local machine | |
|---|---|---|
| Uptime | Runs 24/7, even when your laptop is closed | Stops when you sleep, travel or reboot |
| Network | 10 Gbps port with a static IP | Home connection, often behind NAT |
| Isolation | Separate server, away from your personal files | Shares the machine with everything else you do |
| Webhooks and APIs | Reachable from anywhere for callbacks | Needs tunnels or port forwarding |
| Scaling | Upgrade CPU and memory in place | Limited to the hardware you own |
| Cost | Flat monthly price | Power, wear and your time |
Real feedback from people who use NodexVPS every day.
“We migrated 45 microservices from AWS to NodexVPS Frankfurt nodes. Our monthly cloud bill dropped by 72% while API latency went from 48ms down to 14ms thanks to the dedicated NVMe drives and AMD EPYC silicon.”
“The Terraform provider and sub-30 second provisioning speed made infrastructure automation effortless. No noisy neighbors, true KVM isolation, and zero arbitrary bandwidth throttles.”
“Running low-latency MetaTrader 5 and automated trading bots on Windows Server in London. 99.99% uptime has been rock solid over 9 straight months. Ping to major broker ECNs is under 2ms.”
Fast hardware, fair prices and real support.
Generate millions of tokens per day at a predictable flat monthly price. Never worry about surprise per-token API invoices or rate limits.
Your prompts, documents, internal corporate memos, and embeddings stay strictly in RAM and encrypted NVMe on your dedicated server.
Open WebUI embeds PDFs, Word documents, text files, and markdown using local embedding models for grounded factual question answering.
Expose http://your-vps:11434/v1 to seamlessly integrate with Cursor, Claude Code, LibreChat, Continue.dev, and custom Python agent scripts.
Built for developers, businesses and anyone who needs a reliable server.
Chat with models about internal documents while prompts and uploads stay on the server you control.
Point Cursor, Continue.dev or your own scripts at Ollama's OpenAI-compatible endpoint instead of a paid API.
Upload PDFs, Word files and notes to Open WebUI and ask questions grounded in those documents.
Give staff separate Open WebUI accounts with roles, all served from a single Ollama instance.
Pull DeepSeek-R1, Llama 3.3, Mistral or Qwen 2.5 and compare their answers to the same prompts.
Run heavy prompt volumes at a flat monthly price with no per-token billing.
A few simple checks that save time later.
A quantised model must fit in memory with room for its context. Size the plan for the biggest model you will load.
Each pulled model takes up NVMe space. Budget for the models you keep plus Open WebUI uploads and embeddings.
Ollama's API has no built-in login. Bind it to localhost and reach it remotely through an SSH tunnel or an authenticated proxy.
The first Open WebUI account becomes the admin. Create it straight after deploying, then decide who else may sign up.
Back up the Open WebUI data volume for chats, users and documents. Models can be pulled again, so they matter less.
Watch CPU and memory while models generate so you know when longer contexts or more users need a bigger plan.
Pick the setup that fits your project.
Run Ollama and Open WebUI as two services with named volumes for models and chat data, linked on a private Docker network.
Install Ollama with its install script as a systemd service and run Open WebUI in a container that connects to it.
Keep Ollama on a high-memory server and run Open WebUI on a smaller one that reaches it over a firewalled port.
From a fresh server to a running app.
Log in over SSH and bring the base system up to date.
Install Docker Engine and the Compose plugin from the official script.
Allow SSH and HTTPS only, then turn the firewall on.
Put your compose file in a project folder and start it in the background.
Launch popular AI tools without manual setup.
24/7 autonomous customer service & conversational bots with zero downtime
Private local LLM inference, document RAG & zero token fees
Hierarchical agent swarms & stateful LangGraph workflows
Automate workflows with native LangChain nodes & persistent webhooks
Dedicated VPS hosting for AI agents
Persistent infrastructure for agent workloads
Browse NodexVPS AI VPS setup guides
Chatwoot, Typebot, Telegram & WhatsApp webhook daemons with Redis session cache
Private local LLM inference & document RAG hub
Hierarchical swarms & LangGraph state checkpointers
Native LangChain nodes & persistent webhooks
Remote VPS workspaces for Claude Code
Always-on hosting for Hermes Agent workflows
VPS hosting for NemoClaw experiments
Always-on VPS hosting for OpenClaw assistants
VPS hosting for multi-agent company workflows
Lightweight VPS hosting for PicoClaw agents
Deploy 24/7 resilient customer support bots with Docker, Redis & webhooks
Deploy private Ollama inference with Open WebUI & document RAG
Deploy CrewAI swarms with LangGraph state checkpointers
Deploy n8n with Flowise AI & persistent webhooks
Set up Claude Code on a self-managed VPS
Run Hermes Agent on your own VPS
Host NemoClaw with persistent server access
Run OpenClaw on NodexVPS VPS infrastructure
Host Paperclip for persistent agent workflows
Run lightweight PicoClaw agents on a VPS
Save 50% with yearly billing. No contract, cancel any time.
Quick answers to common questions.
Yes! Modern GGUF quantized models running via llama.cpp inside Ollama take full advantage of AMD EPYC AVX-512 vector instructions. 4-bit and 8-bit quantized models deliver fast interactive typing speeds (15-30+ tokens/sec) on multithreaded vCPU cores at a fraction of expensive GPU rental costs.
Both services run in isolated Docker containers bridged via Docker network. Open WebUI communicates with Ollama over internal container networking with zero public internet exposure required.
Yes. Ollama natively exposes an OpenAI-compatible /v1 API. You can route external agents and IDEs to your VPS endpoint protected by Caddy TLS and an API key.
boots.
Yes. You can move to a larger plan from the control panel. Your data and IP address stay the same.
No. Plans are billed monthly or yearly with no setup fee, and you can cancel at any time.
Ubuntu, Debian, AlmaLinux, Rocky Linux, CentOS Stream, Arch Linux, openSUSE, RouterOS and Windows Server, or any OS from your own ISO.
Yes. Every server sits behind always-on DDoS filtering at no extra cost.
Find the right server for your project.