Your Own AI

Components

What the app can get for this computer - as opposed to add-ons for your AIs. Every component downloads only when you ask and says how big it is first; where it lives and how much space it takes are yours to manage in Settings. The app stays small until you want more.

Download the app

Engines

The bundled engine (Vulkan on Windows and Linux, Metal on Apple Silicon) runs any model on any GPU. Optional engines help specific hardware.

CUDA engine

~800 MB

For NVIDIA GPUs: much faster reading of long questions and documents. Offered after your first answer, never downloaded without asking.

MLX engine (preview)

~60 MB

Apple Silicon: an alternative engine on the same models, for Macs where it measures faster.

Projects

Your AIs working in a folder with you.

Your Own AI Build

~50 MB

Lets an AI read, edit and run in a project folder - every step approved by you, every step in your records. Installs the first time you open a project.

Memory, vision and reading

Small models that run entirely on your device.

Memory and smart routing

~70 MB

Lets each AI recall relevant past conversations and notes, and spots when a question needs an online model.

On-device extraction and smart modes

~2.2 GB

Learns your memory facts and notices when a question wants a report or code - kept on your device even while chatting with online AIs.

Vision

~0.8 to 2.1 GB per model

Lets a downloaded chat model see images you attach - screenshots, charts, documents. One per model family.

Scanned-document OCR

~30 MB

Reads text from scanned or photographed PDFs so they can be summarized, searched and cited like any other document.

Programs tools need

Some tools in Add-ons ask for one of these. The app checks, and installs them where it can - uv with its official installer, Git, Node.js, Python and Docker through your system's own package manager - showing the exact command before it runs.

uv

~20 MB

Runs Python tool servers such as Blender's and FreeCAD's.

Git

varies

Fetches a tool's source when a listing points at a repository.

Node.js

varies

Runs JavaScript tool servers such as Playwright and the 3D printer server.

Python and Docker

varies

For tools that run on Python directly, or ship as containers.