Open WebUI Hosting on a Private VPS

Open WebUI is a self-hosted chat interface that looks and works much like ChatGPT: conversations, model switching, document uploads, prompts you can share with your team, and an admin panel for users and permissions. Choose it with a plan, pay, and we build the VPS (KVM, NVMe, Ubuntu 24.04 LTS) and install Open WebUI, currently version 0.12.4, in about five minutes after boot. You get:

• An HTTPS address like yourname-x7k2.devoster.app with a Let's Encrypt certificate.
• Your admin account already created with your Devoster account email and a generated password you can reveal or reset in the dashboard. The first-run screen, where the first visitor becomes admin, is never left open to the internet.
• Sign-ups closed by default. You add people from Open WebUI's admin panel and decide what each role can use.
• Managed updates. Open WebUI moves fast; we test each version, apply it in a nightly window you choose (or on Update now), snapshot the VPS first and roll back automatically if it doesn't come back healthy. Auto-update can be turned off.

The app is free; you pay for the VPS plan (10% off yearly). See all the apps that install this way on the One-Click Apps page.

The one-click Open WebUI does not include a model. It is the interface, and the models come from providers you already pay: add an OpenAI API key, or any service that offers an OpenAI-compatible endpoint (OpenRouter and many others do), under Admin Panel → Settings → Connections. You can add several connections and let users pick models from all of them.

Why no local models? Devoster VPS plans are CPU-only, with no GPU. Language models that are useful for daily work need a GPU or a lot of memory bandwidth to answer at a comfortable speed, so we don't bundle Ollama or a local model. You have root, so nothing stops you from experimenting, but expect small models and slow answers.

That makes the cost model simple: the VPS is a flat monthly price, and model usage is billed by your provider per token. For a team whose usage is uneven, paying per token through one shared interface can cost less than a chat subscription for every person; for a few heavy users it may not. Check your provider's pricing against how your team actually works.

Self-hosting Open WebUI moves some data under your control, not all of it, and it is worth being precise:

• On your VPS: user accounts, chat history, uploaded documents and their search index, shared prompts and settings. They live in Open WebUI's data directory on a server where you have root, covered by the nightly backups.
• Sent to your model provider: every message and the document passages included with it. That traffic is governed by your provider's API terms, not by where the interface runs.

So the gain is real but specific: one place where your team's conversations are stored, under accounts you control, instead of in each person's consumer chat account. If your data must not leave a particular country, check where your model provider processes requests, and note that Devoster's servers are in Utah, USA.

Because the model runs at your provider, the server only runs the interface, its database and document processing. Open WebUI uses about 0.7 GB of RAM at idle and its image takes about 6.4 GB of disk, which fits comfortably even on Micro's 50 GB.

Document uploads are where the CPU and memory go. By default Open WebUI turns documents into searchable embeddings on the server itself, which is fine on these plans but takes time and memory for large files. If uploads feel slow, switch the embedding engine to your API provider in the admin settings, or move up a plan from the panel.

To give your team a memorable address such as chat.example.com, add it in the Devoster dashboard: a CNAME to your devoster.app name for a subdomain (recommended) or an A record to the server IP for a root domain. The panel verifies DNS and issues the certificate automatically. From the same dashboard you can open the app, reveal or reset the admin password, restart, stop or start Open WebUI, read its logs and run an update.

We manage Open WebUI's installation and updates. The server itself is an unmanaged VPS: you have full root, and your API keys, your users and what you connect the app to are your responsibility. We don't support using Open WebUI itself; its documentation covers that. A few habits worth keeping:

• Set spending limits on your provider account, since every user's messages are billed to your key.
• Give users the "user" role and keep admin to the people who manage connections.
• Read Open WebUI's licence before re-branding it for clients: it requires keeping the Open WebUI branding.

For the server side, our VPS hardening checklist covers SSH and the firewall.