Install-and-go local AI

macOS installers bundle the Ollama runtime. Windows and Linux use slim builds — the setup wizard Install Ollama on first launch (internet required). Open Neural Junkie, pull a default model once, and your agents run on your machine.

Local AI in Neural Junkie: bundled runtime and model library

What happens on first launch

Install from Downloads (.dmg on macOS, .msi or setup .exe on Windows, .deb on Linux).
On macOS, the app auto-starts ollama serve with models in app data. On Windows/Linux, use the wizard Install Ollama first.
The setup wizard offers local or cloud — local pulls a suggested default model once (internet required that one time).
Chat with Assistant, spin up custom experts, or bind Slack — same hub, local inference.

Why it matters

Fewer steps on macOS — runtime ships in the .dmg; Windows/Linux use one-click wizard install.
Privacy-friendly default — run assistants and specialists on hardware you control; add cloud keys only when you want burst capacity.
Model library still built in — open the toolbar Model library (⇧⌘M) to browse curated Ollama and Hugging Face entries after the first pull. See Model App Store guide.

Local image generation (beta.4+)

Generate images with local Ollama models — no cloud API required. Configure under Settings → Image generation.

Limits to know

Installers bundle the runtime, not every model — first model pull can be several GB.
macOS builds are ad-hoc signed, not notarized — use Right-click → Open if Gatekeeper warns.
Cloud providers (Claude, OpenAI-compatible APIs) remain optional in Settings.