OS·WholeTech
OS·WholeTech / Windows
🪟 Windows 10 & 11

The AI stack on Windows.

Everything you need to run modern AI tools on a Windows PC — Claude Code, Codex, Gemini, Tailscale, and Ollama. Do it once, then repeat on every machine.

Works on Windows 10 (22H2) and Windows 11. You'll mostly be typing into a black window called a terminal — don't worry, every command is here to copy.

Before you start

What you'll need

1

Terminal & package manager

Windows Terminal is the modern black command window. winget is Windows' built-in app installer — it downloads and installs software with one line, so you don't hunt for download buttons.

Why this first: every tool below installs through the terminal. Get comfortable here and the rest is copy-paste.

Open a terminal

Press the Windows key, type terminal, right-click Windows Terminal and choose Run as administrator. On Windows 11 it's already installed; on Windows 10 get it free from the Microsoft Store ("Windows Terminal").

Check winget is there
winget --version
✓ Working when: it prints a version like v1.x. If it says "not recognized," update "App Installer" from the Microsoft Store, then reopen the terminal.
Install Node.js (the engine three of the AI tools run on)
winget install OpenJS.NodeJS.LTS

Close the terminal and open a new one after this so it picks up Node. Then confirm:

node --version
npm --version
✓ Working when: both print version numbers (e.g. v22.x.x and 10.x.x).
Optional but recommended: WSL2 (a real Linux inside Windows)

Many AI tools feel more at home on Linux. WSL2 gives you Ubuntu in a window without touching your Windows install. You don't need it to start, but it's the smoothest path long-term.

wsl --install
💡Tip: If you install WSL, you can instead follow the Linux guide inside the Ubuntu window — many people prefer that for Claude Code and Ollama. For now, the rest of this page stays on plain Windows.
2

Claude Code

Anthropic's AI coding agent — the same kind of assistant that built this page. It lives in your terminal, reads and edits real files, and runs commands for you.

Why you want it: it does actual multi-step work (build a site, fix a script, organize files), not just chat.

Install

The simplest, most reliable way on Windows is through npm (which you just installed with Node):

npm install -g @anthropic-ai/claude-code

There's also a native one-line installer if you prefer; run it in PowerShell:

irm https://claude.ai/install.ps1 | iex
Start it & log in

Go into any folder you want to work in, then launch:

claude

The first run opens your browser to sign in to your Anthropic account (a Claude subscription or API credit). After that, just type what you want in plain English.

✓ Working when: running claude --version prints a version, and typing claude drops you into a prompt that greets you.
📘Go deeper: the full Claude playbook — settings, skills, MCP servers — lives at claude.wholetech.com.
3

Codex

OpenAI's command-line coding agent. Same idea as Claude Code, different brain — handy as a second opinion or when you've used up one tool's quota.

Why you want it: variety. Different models are stronger at different things; having both means you're never stuck.

Install
npm install -g @openai/codex
Start it & log in
codex

First run lets you sign in with your ChatGPT/OpenAI account (or an API key). Then describe what you want, same as Claude.

✓ Working when: codex --version prints a version and codex opens its prompt.
4

Gemini CLI

Google's command-line AI agent. Notable for a very generous free tier and an enormous memory (context window) for long documents.

Why you want it: free headroom. Great for big, sprawling tasks before you spend on the others.

Install
npm install -g @google/gemini-cli
Start it & log in
gemini

First run opens your browser to sign in with a Google account. The free tier is large — you likely won't pay anything to start.

✓ Working when: gemini --version prints a version and gemini opens its prompt.

Hermes Agent bonus · 4th agent

A coding agent from Nous Research — "the agent that grows with you." Same idea as Claude Code, Codex, and Gemini, with two twists: it's self-improving (it learns your preferences over time), and it's model-agnostic — you point it at whichever brain you want (Nous, OpenAI, Anthropic, OpenRouter, and more).

Why you'd add it: it's a fourth tool in the rotation and the most model-flexible of the bunch — handy for staying un-locked-in (see the future-proofing principles).

Heads-up: native Windows support is early beta. Nous Research recommends installing it inside WSL2 (the Ubuntu-in-Windows you can set up in Step 1 with wsl --install). So this one lives in your WSL2 window, not plain PowerShell.
Recommended: install inside WSL2

Open your Ubuntu (WSL2) window, then run the one-line installer — it sets up everything it needs (Python, Node.js, ripgrep, ffmpeg):

curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash
Alternative: install with pip (if you already have Python)
pip install hermes-agent
hermes postinstall

hermes postinstall pulls in the extra pieces (Node.js, browser, ripgrep, ffmpeg) and runs the setup wizard.

Set it up & start it
hermes setup        # configure it the first time
hermes --tui        # start it (modern terminal UI, recommended)

During setup it asks how you want to sign in — an API key, or an OAuth login via hermes setup --portal. It works with the Nous Portal, OpenAI, Anthropic, OpenRouter, and others, so you can reuse an account you already have.

✓ Working when: hermes --version prints a version, and hermes --tui opens its interface.
📘Official docs: hermes-agent.nousresearch.com/docs. Since you run it inside WSL2, the Linux guide applies to that window too.
5

Tailscale

A private network (a "mesh VPN") that connects all your devices to each other securely — your 12 PCs, your Macs, your phone, your NAS — as if they were in the same room, from anywhere in the world.

Why you want it: this is the glue. Once every device is on Tailscale, you can reach any of them — to share one Ollama for the whole house, or to control a PC from your phone (see the iOS and Android guides).

Install
winget install tailscale.tailscale

Or download the installer directly from tailscale.com/download/windows.

Sign in

After install, a Tailscale icon appears in the system tray (bottom-right). Click it → Log in, and sign in with the same account on every device. Use the same login everywhere so they all join one private network.

✓ Working when: open a terminal and run tailscale status — it lists this machine and any others already signed in, each with a 100.x.y.z address.
💡Naming tip: in the Tailscale admin console you can rename each machine so your fleet is easy to tell apart (e.g. cc-office-pc, hs-mac-mini).
6

Ollama

Runs AI models on your own PC instead of the cloud — free, private, and works offline. Good for chat, summarizing, and coding help without a subscription.

Why you want it: no per-use cost, nothing leaves your machine, and combined with Tailscale (Step 5) one beefy PC can serve models to your phone and every other device.

Install
winget install Ollama.Ollama

Or download from ollama.com/download/windows.

Download and run your first model

Open a new terminal and pull a small, capable model:

ollama run llama3.2

The first run downloads the model (a couple of GB), then you can chat right in the terminal. Type /bye to leave.

✓ Working when: after the download, it gives you a >>> prompt and answers a question you type.
Hardware note: bigger models need more memory. A model needs roughly its size in free RAM (a 4 GB model ≈ 8 GB RAM to be comfortable). A graphics card (NVIDIA GPU) makes it much faster but isn't required. Start with llama3.2 and go bigger only if it's smooth.
🌐Share it across the house: to let other Tailscale devices use this PC's Ollama, set it to listen on the network — in PowerShell: setx OLLAMA_HOST "0.0.0.0:11434", then restart Ollama. Other devices reach it at this PC's Tailscale address on port 11434.
You're done

What you have now

This PC can now run three cloud AI agents, sits on your private Tailscale network, and serves local models with Ollama. Repeat steps 2–6 on your other Windows machines — it goes fast once you've done one.