AI Server Node

At-home AI server with GPUs for local LLMs

Run large language models on hardware inside your house — not only in someone else’s API region. The MD Vault AI Server Node is a purchased GPU node that sits on your LAN next to the gateway, so prompts and context stay under your roof with your personal data.

Local inference List $3,499 · Data + AI Stack $4,799

What ships

Dedicated AI server chassis with GPU capacity sized for home/office local models

Preloaded container path for local LLM runtime (Ollama-class stack) on your LAN

LAN join to the MD Vault Gateway — private DNS name, firewall path, and AI Guard visibility

Power, cooling, and claim path so the node binds to your MD Hub owner account

Operator console hooks so agents and household tools call models on-prem

The gateway remains the data-home edge (files, photos, contacts, DHCP, Wi‑Fi). The AI node is the inference box. Buy either alone or the Data + AI Stack bundle.

Built for private models

GPU inference

Hardware GPUs on the node run quantized and full-weight local models without sending every prompt to a public API.

Stays on the LAN

Talks to phones, PCs, and MD Hub agents over house Wi‑Fi and Ethernet — behind your gateway edge.

Container-ready

Docker path for model runtimes and approved AI services, restartable on your metal.

Watched by AI Guard

When paired with the gateway, Guard can see node health and network paths the same way it watches other vault servers.

What you can run

Class	On this node (typical)
Coding / agent work	14B–32B class coder and general models at Q4-class quant on ~24 GB VRAM path
Chat & reasoning	Household assistants, summarization, and private document Q&A against vault libraries
Embeddings	Local embedding models for search over files and notes you brought home
Out of band	70B+ full Q4 without multi-GPU or heavy offload — ask Contact for multi-GPU quotes

Class

On this node (typical)

Coding / agent work

14B–32B class coder and general models at Q4-class quant on ~24 GB VRAM path

Chat & reasoning

Household assistants, summarization, and private document Q&A against vault libraries

Embeddings

Local embedding models for search over files and notes you brought home

Out of band

70B+ full Q4 without multi-GPU or heavy offload — ask Contact for multi-GPU quotes

AI node

AI Server Node

$3,499

Single GPU-class node for local LLMs. Physical product — not a cloud seat.

Buy AI node

Best stack

Data + AI Stack

$4,799

Gateway + AI Server Node. Data home and private inference together.

Buy stack

Edge

Gateway alone

$1,499

Start with data home and network; add the AI node when you want local models.

Gateway

List prices are USD MSRP for physical goods. GPU SKU generation and VRAM class ship with the order confirmation. Multi-GPU / enterprise nodes quote on Contact.