Canto

What is Canto?

Canto is a private AI notebook for Mac that revolutionizes how you research, draft, and organize information. It runs entirely on your device using local models, ensuring your notes, PDFs, audio, and web content never leave your Mac—except for encrypted iCloud Drive sync, which you control. This means no readable cloud uploads and no server-side AI, giving you complete privacy and data ownership.

Key features include a versatile document system with notes, notebooks, and a new Canvas—an infinite whiteboard where you can spread out ideas in two dimensions. Drop sticky notes, text, shapes, and pull in real note cards and tag pills from your vault. Connect them manually or ask the canvas agent to find related notes and draw labeled connections explaining relationships. This semantic note linking transforms your knowledge base into an interconnected web.

Canto supports code notebooks, PDF research, and integrates with local AI models via a curated Unsloth GGUF lineup. Default is Granite 4.1 3B, with options like Gemma 4, vision-capable Qwen, and a 122B flagship MoE. You can also connect your own models through Ollama or LM Studio. The Turbo engine uses speculative decoding for up to 2× faster text generation without quality loss, with a live tokens-per-second readout. Note: Turbo is text-only and a separate download; Standard remains default.

Use cases include academic research, content creation, project management, and personal knowledge management. Benefit from offline access, no subscription fees, and a one-time upgrade for unlimited AI. Technical details: Metal acceleration, RAM-based model selection (with in-app warnings for low memory), and over 40 keyboard shortcuts for efficient workflow.

Canto is perfect for privacy-conscious professionals, students, and creatives who want powerful AI assistance without compromising data security. Its local-first approach ensures your ideas stay yours.

Key Features

private AI notebook local model processing encrypted iCloud sync infinite canvas whiteboard semantic note linking code notebooks PDF research multi model support speculative decoding

Who is it for?

researchers, writers, students, knowledge workers, privacy, conscious users, Mac users