Open WebUI is an open-source chat front end for AI models. This guide walks through what you actually need for a self-hosted install (Docker, an OpenAI-compatible API or a local Ollama, a pinned secret key, signup lockdown, updates and backups) and compares that honestly with Private AI Chat (Open WebUI) hosted in Germany.
Open WebUI is a self-hosted chat interface: accounts, conversation history, optional document chat and admin controls. It is not a model provider. Inference happens wherever you point it: an OpenAI-compatible API, another remote endpoint, or a local Ollama instance if the box can actually run the model. The project is open source on GitHub; this guide is about self-hosting or ordering a third-party one-click container in Germany, not about selling model tokens.
If you or your users are in the EU, two points matter immediately: prompts, replies and account data are personal data when they relate to identifiable people, and any connected model provider is its own transfer: hosting Open WebUI in Germany does not move OpenAI, Anthropic, Groq or another API vendor into the EU by itself.
Everything below is grounded in the official docs and the published image as read on 6 November 2026 (Open WebUI Docs: Getting Started, Open WebUI Docs: Quick Start, Open WebUI Docs: Updating, open-webui/open-webui (GitHub)). Releases move; check the docs again before you install.
WEBUI_SECRET_KEY. If it is empty, a new key is generated on start: every restart or image update logs everyone out. Generate once, put it in .env, keep it.ENABLE_SIGNUP=false (or the equivalent admin setting) so strangers cannot register on an open URL.Our longer hands-on walkthrough (Compose, API base URL, pinning the secret key, closing signup, and why local Ollama on a tiny VPS disappoints) is in Self-host Open WebUI on a VPS. The short version: install is easy; the pain later is an unpinned secret, open signup on the public internet, or sizing the box as if it could run a useful local model.
If you already run Docker, a reverse proxy and TLS, bringing Open WebUI up is the easy part. Ongoing work decides it: image updates and version pinning, volume backups (chat history and config live in the container volume), certificate renewal, monitoring, rotating API keys, and someone responsible when the UI is down at night. If you enjoy that, self-hosting is the cheapest option. If you would rather not own it, that is what one-click hosting is for.
Full disclosure: this is what we sell. If you want Open WebUI without running a server, we host an isolated instance for you. As listed in the shop on 6 November 2026:
Shop description, verbatim: “Your own private AI chat workspace (Open WebUI) on an isolated instance. You connect your own model provider and key. Runs as a persistent container. Hosted in Germany (EU); memory and CPU are capped by the container runtime.” Region chip: Germany Available. Price shown: € 12.00 / mo. Access details arrive by e-mail after provisioning. Monthly billing in EUR, USD or GBP; pay by card or SEPA bank transfer. All prices are final totals under the German small-business rule (§19 UStG); no VAT is shown.
This app runs in the EU region (Germany) of the overnight.host network. Our VPS, game server, voice and web plans run in the US (Dallas, TX and Charlotte, NC), with more regions planned.
One-click apps sit under a hostname on apps.overnight.host with a real wildcard certificate (as described on the automation / apps overview). You get the app and a URL, not a root shell. Model tokens are never included, same as self-hosting with a remote API.
| Self-host | One-click (overnight.host) | |
|---|---|---|
| Server | Your VPS; size for the UI (~2 GiB floor without local models) | Isolated persistent container in Germany; RAM/CPU capped by the runtime |
| Setup | Compose + proxy + secret key + first signup + close registration | Automatic after payment; access by e-mail |
| TLS + hostname | Your domain, reverse proxy, certificates | Included under apps.overnight.host |
| Model / API | You: OpenAI-compatible provider, Ollama or another endpoint, plus the bill | You still bring provider and key; we do not sell tokens |
| Updates | You: pull/pin image, recreate, backup volume first | Platform managed for you; no published update schedule on the order page |
| Backups | You: Docker volume off-server (history, config, uploads) | Not itemised as a separate backup SLA on the order page: ask by e-mail before ordering if you need written backup terms |
| Root / custom compose | Full control; co-locate Ollama, RAG extras, custom networks | No root shell; no arbitrary host packages or co-located GPU/Ollama on the same box |
| Data location (app) | Wherever your server is | Germany (EU) for the Open WebUI instance |
| Cost | VPS price + your time + model API usage | €12 per month + your model API usage |
This is informational, not legal advice.
Order Private AI Chat (Open WebUI) → · Self-host Open WebUI on a VPS · Hosting Nextcloud · Hosting Ghost CMS
Written by the person who runs overnight.host: a small, honest hosting company on dedicated bare metal. Live status at up.overnight.host.
Yes, you can self-host the open-source software. Model inference is separate: you pay your API provider or run local hardware yourself.
Only if you want local models on the same machine. For a small VPS, pointing Open WebUI at an OpenAI-compatible API is the shape that usually works.
No. The shop product is the Open WebUI instance. You connect your own model provider and key; their usage charges are separate.
No. You get the app, a hostname and a certificate. For root and custom compose, self-host on a VPS.
Yes, the order page for ai-open-webui showed Germany Available and € 12.00 / mo when checked on 6 November 2026. This product is the Germany one-click app; US VPS locations (Dallas, TX and Charlotte, NC) are a separate catalogue.
Your own Open WebUI instance, hosted in Germany, for EUR 12 a month. Bring your own model API key. Prices are final totals; no VAT is shown (§19 UStG).
Order Private AI Chat → Ask a question first