Start with the machine you have

Suggested hardware for Nurse AI OS

The browser-first Nurse AI OS path does not require a new computer. This page covers the optional Hermes desktop path, local folders, and local-model experimentation—so you can choose a tier based on the work you actually intend to do.

See the two setup doors →

No hardware purchase is required to begin. Start with the browser AI and device you already use. Upgrade only when a specific workflow—local models, sustained multitasking, or parallel agents—earns the cost.

Minimum · light desktop use

Enough to begin

Suitable for browser AI, cloud/API model routes, local Nurse AI OS files, and a light Hermes session.

Mac: Apple Silicon M1 or newer (including current M1–M5 Macs) with 8 GB unified memory. The 8 GB MacBook Neo uses an A18 Pro chip and fits this minimum/light-use tier.
Windows: a dual-core 64-bit processor, 8 GB RAM, and a Windows 11-compatible system.
Storage: an SSD is strongly preferred.

Not the local-model tier: 8 GB systems have little headroom for large models, many browser tabs, or parallel agents.

Recommended · everyday flexibility

Comfortable for most people

The practical target for a personal Hermes workspace that can multitask without making memory the constant bottleneck.

Mac: Apple Silicon M1 or newer with 16 GB unified memory. Current M-series Macs may be M1–M5; memory remains the more important planning choice.
Windows: a recent multi-core processor, 16 GB RAM, and Windows 11.
Storage: 512 GB SSD is a comfortable starting point if you expect local models or several project folders.

Best default: choose this tier if you are buying new and want room to learn before deciding whether local models are worth it.

Optional · local models and parallel work

Buy for a specific workload

Local inference and parallel agents are memory- and storage-hungry. The exact model size matters more than a generic “AI computer” label.

Memory: 16 GB is a practical floor; 24–32 GB is more comfortable for multiple local models or heavier multitasking.
Storage: 512 GB or more of fast SSD space leaves room for model files and updates.
Compute: a stronger CPU/GPU can help, but only buy for a named model or workflow that needs it.

Privacy note: local hardware does not automatically make every workflow private. Provider processing, tools, permissions, and the data you enter remain separate decisions.

Mac and Windows, plainly

What to look for

Component	Minimum practical tier	Recommended starting point
Mac processor	Apple Silicon M1 or newer; the MacBook Neo is a separate A18 Pro option	Apple Silicon M1 or newer, including current M1–M5 Macs. Newer generations may improve efficiency, but the memory tier matters more than the number alone.
Mac memory	8 GB unified memory	16 GB unified memory. An 8 GB A18 Pro MacBook Neo can serve as a minimum/light-use machine; it is not the sensible choice for heavy local-model work.
Mac operating system	macOS 12 or newer as a planning floor	Use a current supported macOS release when possible. Confirm the current Hermes support matrix before installing on an older system.
Windows processor	Recent dual-core 64-bit processor	Recent multi-core processor for multitasking, local tools, and longer sessions.
Windows memory	8 GB RAM	16 GB RAM; more if you plan local models or parallel work.
Windows operating system	Windows 11-compatible system	Keep Windows and device drivers current enough for the applications and model tools you choose.
Storage	SSD preferred; keep free space	512 GB SSD if local models are part of the plan. Check each model’s listed download size first.
Internet	Ordinary broadband	Stable broadband or Wi-Fi. Bandwidth matters most when downloading local models or updates.

These are practical planning tiers, not a guarantee that every Hermes feature, model, connector, or local runtime will work on every machine. Check the official Hermes documentation ↗ before installation. For the 8 GB MacBook Neo reference, see Apple’s technical specifications ↗.

The honest cost model

The honest cost of running Hermes with Nurse AI OS

Hermes is free and open-source. Nurse AI OS can begin browser-first without a hardware purchase or a required subscription. The real costs come from what you choose around the system—and each one is optional.

1. Model or API usage

Paid model providers may charge per token, request, or subscription. A free provider route, an existing authorized subscription, or a local model can keep the model bill at zero—but free routes may have limits, changing terms, or weaker capability.

2. Electricity

Cloud and API use add little to the wall-power bill of your laptop. Local models can keep the CPU or GPU working harder, so sustained local inference may use noticeably more electricity.

3. Cloud time

If you choose a cloud runtime, VM, or hosted agent, read its billing policy. Some services meter active work; others also charge for uptime, storage, networking, or reserved capacity. An idle agent is not automatically free.

4. Hardware

Hardware costs money only when you buy or upgrade it. Most people should not upgrade first. Let a specific job—local models, large files, sustained multitasking, or parallel agents—justify the purchase.

Honest summary: a careful browser-first or cloud/API setup can have zero new marginal cost. Costs appear when you deliberately choose paid models, metered cloud resources, higher electricity use, or new hardware.

Questions before you buy

Practical answers

Does disk speed matter?

For cloud and API use, barely. For local models, yes: models are large files, and an SSD makes downloads, loading, updates, and workspace searches much more pleasant than an old spinning drive.

How much disk space do local models need?

They are multi-gigabyte downloads, and sizes vary widely. Check the model’s listed size before downloading it. Keep comfortable headroom if you plan to try several models, retain project files, or keep room for updates.

Does my internet connection matter?

Ordinary broadband is fine for typical API requests, which exchange relatively small amounts of text. The bandwidth-heavy moment is downloading a local model or a large update, usually once per model.

Is 8 GB enough?

It can be enough for browser-first use, cloud/API routes, local files, and light Hermes work. It is not a comfortable target for local models, many browser tabs, or parallel agents. If you are buying new and can choose, 16 GB is the better default.

Can I start on a phone, tablet, or Chromebook?

Yes, for the browser-first Nurse AI OS experience. A supported Mac or Windows desktop or laptop is needed only when you choose the optional Hermes desktop/local-runtime path.

Requirements follow ambition. Start at the minimum tier and lose nothing in the browser-first path. Move up only when local models, privacy preferences, or parallel agents earn their keep. The machine should support the work—not become the excuse to postpone it.

Hardware guidance is informational and changes as operating systems, Hermes releases, model sizes, and providers change. This page is not a clinical, institutional, security, or compatibility certification. Keep all Nurse AI OS and Hermes work no-PHI unless an explicitly authorized environment and accountable governance say otherwise.

Start with the setup guide → Return to Resources