Proxmox + NVIDIA A4000 'AI Basecamp' reachable anywhere via Tailscale
Tailscale's Alex Kretzschmar passes an NVIDIA A4000 through to a NixOS VM on Proxmox, runs Ollama + Open WebUI in Docker, and exposes it privately with Tailscale Serve (TLS, no reverse proxy).
Solo2-10 peopleWriting assistantCodingDocument search / chat with files$1k-3kSensitivity: High (client/confidential)
Hardware
| Item | Approx. price (as of 2026-10-16) |
|---|---|
| NVIDIA RTX A4000 workstation GPU in a Proxmox host | Not stated in source |
| NVIDIA RTX A4000 16GB (Newegg, related current price) | $1,399.00 |
Source prices are as stated by the source. Retailer offers: prices checked 2026-10-16, may change. Some links may be affiliate links; we may earn a commission at no extra cost to you. This never changes what we list.
Software stack
- Proxmox
- NixOS
- Docker
- Ollama
- Open WebUI
- Tailscale Serve
- Ansible
Models used
- Local models such as Llama 4, Gemma 3, DeepSeek (mentioned)
Good at
- Private access from phones/tablets on your tailnet with TLS
- Reproducible (NixOS + Ansible)
- Share access via node sharing
Weak at
- Needs PCIe passthrough-capable hardware and BIOS
Rough total cost
Not stated in source.
Getting started
- Install Proxmox and enable PCIe passthrough for the GPU.
- Create a NixOS VM with Docker, NVIDIA drivers and Tailscale.
- Run Ollama and Open WebUI containers.
- Use Tailscale Serve to publish Open WebUI to your tailnet only.
Source & credit
Self-host a local AI stack and access it from anywhere (Alex Kretzschmar, May 2025) — Tailscale blog
Summarized from the source; quotes are the source author's words. Check the original before buying.