Version 0.1.0 · self-contained installer

Download the AI Oracle desktop agent

One installer. Everything bundled — Python interpreter, model weights, VS Code extension. No setup steps after running it.

We couldn't detect your OS automatically — pick the right artifact below.

Windows .exev0.1.0

macOS .dmg — coming soon

Linux .deb — coming soon

Local-first

Embeddings run on your machine. Your decisions and activity never leave your hardware unless you opt into cloud sync per card.

Zero setup

No Python, no pip install, no PATH edits. The installer ships everything the agent needs — including the MiniLM embedding model.

VS Code, automatic

If you have VS Code on your PATH, the installer registers the AI Oracle extension for you. No second download.

What gets installed

%LocalAppData%\Programs\AI Oracle\

├── AI Oracle.exe — the tray-resident agent

├── oracle.exe — Python core, sentence-transformers, lancedb

├── _internal\ — bundled libraries (~500 MB, mostly torch)

├── models\all-MiniLM-L6-v2\ — embedding model, pre-bundled

└── resources\ai-oracle-vscode-0.1.0.vsix — VS Code extension

The installer also seeds ~/Desktop/AI Oracle Sample Corpus/so first-run UX is “open the folder, start writing” instead of “go read docs.”

First run

1. Run the installer. Windows SmartScreen will warn the first time — click More info → Run anyway. The build is unsigned during early access; we’re working on signing.
2. Wait for it to finish. Per-user install, no UAC prompt. Takes 30 seconds on an SSD.
3. Open the tray icon. Settings → point Watched directory at the seeded sample corpus on your Desktop. Save.
4. Edit a file in that directory. On save, the cascade fires; if your draft conflicts with a card, you’ll see a toast.

Build from source

The desktop agent is open inside the AI Oracle repo. Clone it, install Rust + Node + Python, and run desktop-agent\installer\build-installer.ps1. See the agent’s README.

Report an install issue