Quick Start Guide

Install in four steps

1. Open the download page and choose Windows or macOS. Windows 10 or later (64-bit) and macOS 10.15 or later are supported, and the Mac build runs on both Apple Silicon and Intel.
2. On Windows, run the installer you downloaded and follow the wizard. If SmartScreen warns you, choose More info and then Run anyway.
3. On macOS, open the DMG, drag the Grout Suite into your Applications folder, then right-click the app and choose Open the first time so Gatekeeper lets it through.
4. Launch Grout from the Start menu on Windows, or from Applications on macOS.

Do you need an account?

No account is needed to run the models that live on your own machine — install Grout, download a local model and start working. An account comes in only when you buy a licence or turn on the optional Nexus AI cloud module, which spends Merits.

Pick a size your machine can hold

A local model needs roughly as much free memory as its file is large, plus one to two gigabytes of headroom for the application. Check a model's size before you download it, and compare it against the memory you actually have free rather than the memory installed.

• On 8 GB of RAM, stay with the smaller general-purpose text models and run one at a time.
• On 16 GB, mid-sized models run comfortably with the canvas open alongside.
• Add a vision model only if you will be reading handwriting, photographs or diagrams.
• Add an image model only if you will be generating or editing pictures.

Local models versus the Nexus AI cloud

Downloading a model needs a connection; running it afterwards does not. The Nexus AI module is a separate, opt-in layer: it reaches 300+ models on Grout's cloud GPUs, needs an account and an internet connection, and spends Merits. Every Grout Suite subscription includes 20 Merits, and you can top up from the Merit Store.

Step by step

1. Create a new project and open the canvas.
2. Bring in your material: import a photograph of your notes, paste an image from the clipboard, or sketch straight onto the canvas with the brush tool.
3. Select the image and run the vision action that extracts text. The handwriting comes back as editable text you can place on the canvas.
4. Ask a tutor to work with that text — a summary, a set of practice questions, or a plain-English explanation of the part you got stuck on.
5. Lay the result out: text boxes for the explanation, shapes to group related ideas, and snapping turned on to keep edges aligned.
6. Save the project, then export the canvas as an image when you want to print or share it.

If the result looks wrong

Vision output is only as good as the picture you gave it. Re-shoot the page with even light and no shadow across the writing, crop tightly to the writing itself, and try again before you change any settings.

Save any prompt you would type twice

The same summary format, the same marking style, the same explanation length — anything you ask for more than once belongs in a saved custom prompt rather than being retyped from memory.

Give each subject its own tutor

A tutor carries its own instructions and its own model. Separate tutors for maths and for history keep each one's tone and vocabulary right without you re-explaining the context every session.

Match the model to the job, not to the benchmark

The largest model you can run is rarely the fastest way to finish. A small model that answers in a second beats a large one you sit waiting on, unless the task genuinely needs the extra capability.