Mini PCs
N1522
Intel Twin Lake N150 (4 cores, 4 threads, up to 3.6GHz) with 2x RTL8111H Gigabit LAN (10/100/1000M).
Turnkey AI agent host
N1522 — an always-on home for your AI agents
AI agents earn their keep by never sleeping. The N1522 is a silent, low-power host built to run them around the clock — dual LAN, DDR5, wide-range DC input, and the recovery features a 24/7 server needs.
Fanless9–36V DCDual LAN
The problem
Why a dedicated agent host?
Laptops sleep
Agents that live on a laptop stop when the lid closes or the laptop travels.
Cloud VMs bill every month
An always-on VM is a recurring cost, and your files and credentials live in someone else’s infrastructure.
Desktops are loud and power-hungry
A tower running 24/7 for a few containers wastes power and adds noise.
Agents need isolation
Tool-using agents should run away from your daily machine, credentials, and personal files.
N1522 is a low-power, always-on x86 Linux host for agent runtimes, Docker services, and automations. The model can be cloud or local; the host stays yours.
Why this platform
Fanless & always-on
Silent operation with auto power-on and watchdog for months of unattended headless duty.
Consolidate with containers
x86 runs Proxmox VE and LXC — several isolated assistants on one low-power box.
Dual LAN + serial
Separate management from service networks; COM ports for agents that touch hardware.
Wide 9–36V DC input
Tolerant of closet and UPS power; storage flexibility with dual M.2 plus a 2.5" bay.
Ports for an always-on host
LAN 1
Service network
Agent traffic, APIs, and dashboards.
LAN 2
Management / isolation
Admin access, a separate VLAN, or a link to a firewall such as the N1241.
2× COM + 8× USB
Local devices
Serial sensors, controllers, and USB peripherals for automation that touches hardware.
Key specifications
- Processor
- Intel Twin Lake N150 (4-core)
- Memory
- 1× DDR5 4800 SODIMM, up to 16GB
- Network
- 2× Gigabit LAN
- Storage
- 2× M.2 2280 + 2.5" bay
- Power
- 9–36V DC wide range
- Cooling
- Fanless
All 9 specifications
- Processor
- Intel Twin Lake N150 (4 cores, 4 threads, up to 3.6GHz)
- Memory
- 1x DDR5 4800MHz SODIMM, up to 16GB
- Networking
- 2x RTL8111H Gigabit LAN (10/100/1000M)
- Display outputs
- 2x HDMI (4096×2160@60Hz)
- Storage
- 2x M.2 2280 + 1x 2.5" HDD/SSD
- I/O
- 4x USB 3.2, 4x USB 2.0, 2x COM RS232/RS485
- Dimensions
- 150 × 126.5 × 45mm
- Power input
- 9-36V DC wide range
- Platform features
- Optional CAN bus, GPIO, Wi-Fi/4G module
Software fit
- Ubuntu Server
- Proxmox VE + LXC
- Docker
- Agent frameworks
- Tailscale / WireGuard
Software compatibility is validated for the exact model and release before quotation.
Dedicated host vs the alternatives
The typical picture; individual setups vary.
| Feature | N1522 | Always-on cloud VM | Your laptop or desktop |
|---|---|---|---|
| Running cost | One-time hardware, low power draw | Monthly bill | Your power, but not always on |
| Where files and credentials live | On your hardware | Provider infrastructure | Your daily machine |
| Availability | 24/7 with auto power-on and watchdog | 24/7 | Stops on sleep or travel |
| Isolation | Separate device and network segment | Separate tenant | Shares your daily environment |
| Local models | Light, quantized models only | Depends on the instance | Depends on the hardware |
Prompts and data sent to a cloud model API leave the device according to that provider’s terms.
Where it fits
Always-on agents
Headless host for agent frameworks calling cloud or local models, 24/7.
Assistant consolidation
Run several isolated assistant instances in LXC containers on one machine.
Home / small-office server
A quiet, capable server for automation, dashboards, and light services.
Typical configurations
Starting points technical sales uses when sizing the N1522. Every build is confirmed against the agents, services, and storage you plan to run.
Starter
A personal agent and automations
- 8GB DDR5
- 256GB M.2 SSD
- Ubuntu or Debian with Docker
Pro
Several agents and services
- 16GB DDR5
- 512GB M.2 SSD
- Proxmox VE with LXC, backups to the 2.5" bay
Protected
Business agents on a segmented network
- 16GB DDR5
- 512GB M.2 SSD + 2.5" SSD
- Paired with the N1241 firewall for VLAN isolation
Frequently asked questions
Can it run large language models locally?
It is an excellent agent host and runs small quantized models acceptably. For heavier local inference, size up to a Core-class N3322 or IBOX-3226 — send the model and throughput target.
How many assistants can it host?
With Proxmox and LXC, several light instances consolidate on one box, each isolated with its own resources. State the count and workload for sizing.
Does my data stay local?
Configuration, files, memory stores, logs, and credentials can stay on the N1522. If an agent calls a cloud model API, the prompts and data it sends leave the device according to that provider’s terms.
Does it need a firewall?
If it faces the internet or runs tool-using agents, yes — pair it with the N1241 AI firewall for VLAN isolation and egress control.
Related hardware
Ready to specify the N1522?
Send the configuration, quantity, application, and timing. Technical sales confirms compatibility, lead time, and pricing.