Connecting Ollama

Connecting Ollama (free to run, on your own computer)

Ollama is an optional extra you can connect to GRU953-Studio. This page explains what it is, why you might want it, and exactly how to set it up — in plain steps, with no assumptions about what you already know.

The short version: everything here is optional. Claude stays the default. You never have to touch Ollama at all, and when the studio does want to use it, it always asks you first.

What Ollama is

Ollama is a free tool (no cost) that runs AI models directly on your own computer, with no cloud needed. "A model" here just means the AI itself — the part that reads your words and writes a reply. Normally that runs on a company's servers over the internet; with Ollama, it runs on your machine instead.

Why you might want it

There are two honest reasons.

• It is private. Because the model runs on your own computer, nothing you send it leaves your machine. Nothing travels to a company's servers.
• It is free to run (no cost). Once it's installed, you don't pay per use.

There's an equally honest catch, and it's worth saying plainly: a model running on your own computer is slower and less capable than Claude. So Claude remains the default for GRU953-Studio, and Ollama is only ever an option you can add on top — never a silent swap.

Two ways the studio uses it

GRU953-Studio can use Ollama in two separate ways. Both are optional, and both always ask before doing anything.

1. As an alternative "engine" for an app the studio builds you. If the studio is building you an app that has an AI feature inside it, it can offer Ollama as the engine that powers that feature — so the finished app runs its AI on your (or your users') own computer instead of the cloud. This is always offered as a choice; Claude stays the default engine unless you pick Ollama.
2. As a free (no cost) second opinion for the team's own work. Any of the studio's specialist roles can ask a local Ollama model to look over their work — a genuinely independent, private check that can sometimes spot things a single reviewer misses. It's a free (no cost), local sanity-check, never a replacement for the team's own review.

The golden rule: it always asks first

This is the most important thing on the page.

GRU953-Studio will always ask your permission before it installs Ollama, and before it downloads any model — every single time, with no exceptions. Installing software and downloading a model are real changes to your own computer, so each one gets its own fresh, plain-English question with the size and cost spelled out. Saying yes once does not give a blanket yes to everything after; the next install or download asks again.

If you say no, nothing is installed or downloaded, and the studio simply carries on with Claude.

Not available on Claude Code on the web

Ollama runs a model on a local computer, so it needs a local computer to run on. That means it is not available when you use Claude Code on the web (or any other setup where the work happens on remote servers rather than your own machine).

You don't need to do anything about this. On those sessions the Ollama features simply switch themselves off and leave a short note explaining why. It's treated as normal, not as an error — the team's own review still stands, and any built app that offered Ollama falls back to its other engine.

Before you start

A few things worth knowing before the first install:

• You only need Ollama if you actually want one of the two uses above. Most projects never need it.
• The studio can install it for you (after you say yes), so you usually don't have to type anything. The steps below are here so you understand what's happening, and so you can install it yourself if you'd rather.
• GRU953-Studio does not bundle or ship any Ollama software of its own. It simply talks to the official Ollama that lives on your computer. You get Ollama from its makers, at ollama.com.

Installing Ollama, step by step

There are two ways to install Ollama on each system: a one-line command, or a normal installer file you download and open like any other app. Pick whichever feels more comfortable.

A "command" below means a line of text you run in a terminal — the plain text window on your computer where you type instructions. If that sounds unfamiliar, the download-and-open option is the friendlier choice.

On a Mac (macOS)

To use the one-line command:

1. Open the Terminal app (press Command and Space, type Terminal, and press Return).
2. Type or paste this line, then press Return:

   curl -fsSL https://ollama.com/install.sh | sh
   

3. Wait for it to finish.

To use the installer file instead:

1. In your web browser, go to ollama.com/download.
2. Download the macOS installer (a .dmg file).
3. Open the downloaded file and follow the on-screen prompts, exactly as you would for any other Mac app.

On Windows

To use the one-line command:

1. Open PowerShell (press the Windows key, type PowerShell, and press Enter).
2. Type or paste this line, then press Enter:

   irm https://ollama.com/install.ps1 | iex
   

3. Wait for it to finish.

To use the installer file instead:

1. In your web browser, go to ollama.com/download.
2. Download the Windows installer (an .exe file).
3. Open the downloaded file and follow the on-screen prompts.

On Linux

1. Open your terminal.
2. Type or paste this line, then press Enter:

   curl -fsSL https://ollama.com/install.sh | sh
   

3. Wait for it to finish.

How much space it needs

It's fair to want to know what this costs in disk space before you agree. Here are the honest figures.

What	How much space
The Ollama install itself	about 4 GB of disk space
Each individual model	roughly 1 GB to well over 100 GB

Models vary enormously in size — small ones are around 1 GB, while the largest run to more than 100 GB. The studio tells you the size of a specific model before it downloads it, so you can decide with the real number in front of you.

One thing we will not do is invent a figure: Ollama does not publish an official table of how much memory (RAM — your computer's short-term working memory) each model needs to run. Rather than guess, we say so plainly. If you're unsure whether your computer can comfortably run a particular model, that's a good moment to ask before downloading.

What happens after you say yes

Once Ollama is installed and running on your computer, there's nothing to set up by hand.

GRU953-Studio talks to Ollama at an address on your own machine: http://localhost:11434. "localhost" simply means "this computer" — so that address points at the Ollama running right there on your own machine, not anywhere on the internet. You don't need to type that address, edit any settings, or configure anything. Beyond saying yes to the install and to any model download, it just works.

What GRU953-Studio does not do

To keep everything clear and honest:

• It does not install Ollama, or download any model, without an explicit fresh "yes" from you for that specific action.
• It does not make Ollama the default. Claude stays the default; Ollama is only ever an option you choose.
• It does not ship any Ollama software of its own — it uses the official Ollama on your computer.
• It does not treat a local model's answer as the final word. Anything the local model reports is checked before it's acted on.

Where next

• New to the studio? Start with Installation.
• Want a paid cloud option for generated images, audio or video instead? See Connecting Gemini.
• Curious who does the work behind the scenes? Meet The team.
• Something not working? Try Troubleshooting or the FAQ.

The full GRU953-Studio project lives at https://github.com/GRU-953/GRU953-Studio, with the main website at https://gru-953.github.io/GRU953-Studio/.