Run free, private AI on your PC with Ollama and OpenCode

Kevin installs Ollama on Windows, runs Qwen 3.5 4B locally with no account, chats, analyzes a customer-feedback PDF offline, then connects it to OpenCode so the local model builds a cookie-shop website as real files.

Kevin Stratvert · YouTube · published Nov 5, 2026 · Software & DevOps

Run free, private AI on your PC with Ollama and OpenCode — video by Kevin Stratvert▶ Watch on YouTube
Prompt
Create a website for the Kevin Cookie Company, complete with a cookie menu, prices, and a contact section.

Visual breakdown

Install Ollama, pick a small local model like Qwen 3.5 4B, and you can chat, analyze documents and even build a website with OpenCode, all free, offline and with your data staying on your computer.

The problem: Cloud AI needs accounts, subscriptions and sending your files to someone else's servers.

From install to a website built by a local model

  1. Install OllamaDownload the Windows installer from the site and choose 'no thanks, I'll use Ollama locally'
  2. Pick a local modelDownload-icon models run on your PC; Qwen 3.5 4B is the small starter
  3. Chat and documentsPrompt it, ask follow-ups, drag in a customer-feedback PDF
  4. Connect OpenCodeOpenCode auto-detects the models already downloaded in Ollama
  5. Local model builds filesOpenCode gives Qwen the tools to create and edit files in a chosen folder
  6. Working websiteAn index.html for the cookie shop, created on disk at no AI cost
InputsSetup / manualAI modelResult

Do it yourself

Everything below runs on Windows 11 in the video; Ollama also supports Mac and Linux.

  1. Download and install Ollama, skipping account creation
  2. Open a new chat and choose Qwen 3.5 4B from the model dropdown
  3. Send a prompt; the model downloads on first use
  4. Drag in a PDF and ask for the top complaints and one fix each
  5. Install the OpenCode desktop app and pick your Ollama model
  6. Ask it to build a website and open the generated index.html

Good to know

  • Models with a cloud icon need an account; download-icon models run locally.
  • Smaller models are faster and fit more PCs but may struggle with complex tasks.
  • Larger models like Gemma 4 need more memory and processing power.
  • Once a model is downloaded it works without an internet connection.

Distilled from the video’s captions

How to try it

  1. Open Ollama.
  2. Copy the prompt and fill in the [placeholders].
  3. Watch Kevin Stratvert's video for the full walkthrough and setup.

Related use cases

Share