Changelog.

✨ New Features

Drag & Drop File Attachments

You can now drag files directly from your computer into any note! Simply drag PDFs, images, documents, or other files from Finder/Explorer and drop them anywhere in the editor. Each file gets its own attachment line, and PDFs/documents automatically have their text extracted for AI analysis.

Supported file types:

• PDFs - Full text extracted and saved for AI
• Images (PNG, JPG, GIF, WebP)
• Word documents (DOCX)
• Text files (TXT, MD, JSON, CSV, XML)
• Code files (.js, .ts, .py, etc.)

Writing Style Profile Toggle

Added a new toggle in Memory Links settings to control whether Canto analyzes your writing style. When enabled, Canto studies your notes to make AI suggestions match your tone and vocabulary. When disabled, no analysis jobs run, saving system resources.

Where to find it: Settings → Memory Links → Writing Style Profile toggle (below Related Notes Sidebar)

Read all the notes

🔧 Improvements

Smarter AI Attachment Reading

When you attach a large PDF (like a 74-page study guide), Canto now stores the full text in a separate file rather than cramming it into your note. This keeps your notes fast and responsive, while still giving the AI access to the complete document content when you ask about it.

More Accurate Context Usage Display

The token usage display at the bottom of AI chats now correctly counts the content from file attachments. Previously, it only showed the chat message tokens - now it includes the PDF/document text that gets sent to the AI.

Better Model Management

• Fixed: Downloaded models now correctly show a trash/delete button instead of the download button
• Fixed: Loading an already-cached model now shows "Loading" instead of "Downloading"
• Added: Newly downloaded models immediately appear as cached in the model list

🛠 Technical Improvements

Behind-the-Scenes Architecture

• Moved attachment text storage from note HTML to companion files for better performance
• Improved IPC communication for file operations
• Enhanced AI tool instructions for more reliable attachment handling

📝 Notes

This release focuses on making file attachments more powerful and easier to use. The drag-and-drop feature should make adding study materials, reference documents, and images to your notes much faster. The Writing Style Profile toggle gives you more control over background processing.

As always, everything runs locally on your device - your files and notes never leave your Mac.

v0.1.0-alpha.3 → February 8, 2026

✨ New Features

AI Can Now Search Your Notes The Agent Chat can now find and read your existing notes. Ask things like:

• "Find notes about my Q4 planning"
• "What do I have about Project Aurora?"
• "Summarize everything I've written about budgets"

The AI searches through all your notes and can pull up relevant content to answer your questions.

AI Can Read Attached Files Drop a PDF, Word document, or text file into a note, and the AI can now read and summarize it:

• "Summarize the attached PDF"
• "What are the key points in this document?"
• "Extract the action items from the report"

Supported file types: PDF, DOCX, TXT, MD, HTML, JSON, CSV, and code files.

Automatic Text Extraction When you attach a text-readable file (like a PDF), Canto automatically extracts the text in the background. You'll see a status like "✅ Text extracted (1,247 words, 12 pages)" on the attachment.

🐛 Bug Fixes

Fixed: AI Chat No Longer Interrupted Fixed an issue where the writing style analyzer would sometimes interrupt your active AI conversations. The chat should now work smoothly without unexpected stops.

---

How to Use the New Features

Searching Notes:

1. Open Agent Chat (Cmd+J)
2. Type: "Find notes about [topic]"
3. The AI will search and show you relevant notes
4. Then ask: "Read the first one" or "Summarize what you found"

Working with Attachments:

1. Attach a PDF or document to any note
2. Ask the AI: "Summarize the attached file"
3. The AI will extract the text and summarize it for you

Note: For PDFs with images only (scanned documents), text extraction won't work. The file needs to have actual selectable text.

v0.1.0-alpha.2 → February 7, 2026

✨ New Features

AI That Writes Like You Canto now learns your unique writing style by analyzing your existing notes. When you use AI autocomplete, it will match your tone, vocabulary, and sentence patterns—making suggestions feel like they came from you, not a robot. You can view your style profile in Settings → Memory Links, and regenerate it anytime your writing evolves.

Smoother AI Responses AI-generated text now streams in more smoothly, with a visual cursor that shows you exactly where new content is being added. No more jarring jumps or flickering—just a natural, readable flow as Canto thinks and writes alongside you.

Better Code Support Running TypeScript code blocks in your notes is now more reliable, with improved error messages if something goes wrong.

Release Date: February 6, 2026
Status: Alpha (Pre-release)

What is Canto?

Canto is your private AI notebook — a desktop app for macOS that combines rich writing, code notebooks, and unlimited local AI. Everything runs on your device. Nothing ever leaves your Mac.

Canto is the successor to Muse, rebuilt with a new name and expanded vision.

Highlights

• Memory Links — AI-powered knowledge graph that automatically discovers connections between your notes as you write
• AI Everywhere — Ask AI (Cmd+K), Agent Chat (Cmd+J), Selection AI, and real-time autocomplete, all running locally
• Code Notebooks — Run Python (via Pyodide), JavaScript, and TypeScript in Jupyter-style cells alongside your notes
• 100% Private — All AI models run locally via WebGPU. Notes stored in encrypted SQLite. Zero cloud uploads.
• 6 Local AI Models — Phi-3.5 Mini, Mistral 7B, Llama 3.1 8B, Qwen 3 8B, and Gemma 2 9B
• Rich Editor — WikiLinks, backlinks, Mermaid diagrams, LaTeX math, split view, slash commands, and more

System Requirements

• macOS 14 (Sonoma) or later
• Apple Silicon (M1/M2/M3/M4)
• WebGPU support
• 8GB+ RAM recommended

Download

Available on the Canto Releases page.

License

One-time purchase — $24.99 for 1 device with lifetime updates.