Machine AI isn't generic advice. It reads your real scan results — actual GB found, specific categories, your disk usage — and gives advice tailored to your specific Mac. Runs 100% locally via Ollama. Zero data ever leaves your machine. Included in the $19.99 one-time purchase.
macOS 13+ · Apple Silicon & Intel · ~1.3 GB model download, one-time
Most AI assistants give generic advice: "try clearing your cache" or "check your storage." Machine AI is different because it has actual context. Before every response, it receives a structured summary of your real scan results:
So instead of "you might have cache files," it says "you have 8.4 GB in app caches — the biggest chunk is likely browser caches. I'd start with Chrome and Firefox." Specific to your Mac, right now.
Machine AI runs via Ollama, an open-source local model runner. The model (llama3.2:1b, ~1.3 GB) downloads once and runs locally after that. Every conversation, every scan result, every question you ask — it all stays on your Mac. No API key, no cloud subscription, no data collection. The AI doesn't even have a way to send data out.
Setting up a local LLM usually requires installing Homebrew, then Ollama, then pulling models in the terminal, then configuring the API. CleanMachine's Machine AI does all of this automatically in a zero-terminal setup wizard:
Future launches skip all of this — the model is already present and Ollama starts in seconds.
The chat interface looks and works like Messages: your questions appear on the right, AI responses on the left. Responses stream token-by-token as they're generated — no waiting for a long response to finish loading. Three animated dots show while the first token is being processed.
When Machine AI first becomes ready (after setup), it automatically reads your current scan state and generates a 2–3 sentence plain-English summary of the top cleaning opportunities on your specific Mac. No prompting required — it greets you with useful information.
Ask "what should I clean first?" and Machine AI has the actual Smart Scan totals — broken down by category, with real GB numbers from your Mac.
"I have 847 duplicate files — what are those likely to be?" Machine AI explains the common sources and which categories are typically safe to bulk-remove.
Ask about specific large files, when to delete old iOS backups, or whether it's safe to clear Xcode DerivedData — Machine AI gives direct, specific answers.
Ask about privacy traces, what "app state" means, or whether you should use Cave Mode — personalized to what your Privacy scan actually found.
Download CleanMachine free to explore the interface. Machine AI and all cleaning features unlock for the one-time $19.99 purchase — no subscription, no ongoing cost, fully private.
macOS 13+ · ~1.3 GB model download (one-time) · No subscription · 30-day money-back guarantee
The llama3.2:1b model is approximately 1.3 GB. It downloads once and lives in CleanMachine's Application Support directory. There's no ongoing storage growth — the model file doesn't change unless you update it.
Yes, completely. After the one-time model download, Machine AI works with no internet connection. Ollama runs as a local subprocess — no API calls, no external services.
llama3.2:1b is a 1-billion-parameter model — specifically chosen to run fast on consumer hardware. On Apple Silicon (M1 or newer), responses stream smoothly. On older Intel Macs it's slower but functional. The model runs on Metal (GPU acceleration) where available.
The current version uses llama3.2:1b specifically, selected for speed and reliability on consumer hardware. Support for other models through the Ollama API is planned for a future update.