Frequently asked questions

Nexus Note is a Mac-native, local-first personal AI system. It ingests your documents, calendar, email, tasks, meeting recordings, and other information sources, indexes them locally, and lets you query and act on them using a local language model.

Think of it as a private intelligence layer over your entire digital life—entirely on your machine, under your control.

No. Nexus Note is not a new place to write. It's a system that sits above the tools you already use—connecting documents, time, tasks, and media into a unified, queryable knowledge base.

You keep using whatever note-taking app, calendar, or email client you prefer. Nexus Note ingests from them and makes the collective knowledge searchable and actionable.

Three things:

• Local: Everything runs on your Mac. Your data never goes to OpenAI, Anthropic, or any other cloud service.
• Grounded: Answers come from your actual notes with citations. No hallucinations about things you never wrote.
• Persistent: Your knowledge base grows over time. It's not a blank-slate conversation—it's an intelligence layer that knows your context.

We're currently in development with early access planned for 2025. Join the waitlist to be notified when spots open up.

Yes. All processing happens locally on your Mac. Your notes, embeddings, search queries, and AI conversations never leave your machine. There's no cloud sync, no telemetry, no "anonymous" data collection.

We can't see your data because we never have access to it.

No. Nexus Note works entirely offline. The AI models run locally, and all your data is stored on your machine. You can use it on a plane, in a secure facility, or anywhere else without connectivity.

Everything is stored in a local SQLite database and vector store in your Mac's file system. You can find the files, back them up, move them, or delete them just like any other files on your computer.

Yes. You can export your entire knowledge base as JSON, Markdown, or a raw SQLite database at any time. Your data is never locked in.

Nexus Note requires a Mac with Apple Silicon (M1 or later) for optimal performance. We leverage Metal and MLX for fast local inference. Intel Macs may be supported with reduced performance.

Minimum 16GB RAM recommended. Storage requirements depend on your knowledge base size—expect roughly 1-2GB per 100,000 notes indexed.

Nexus Note uses a combination of local models:

• Embedding models: For semantic search and similarity matching
• Language models: For answering questions, summarization, and drafting

All models run locally using optimized inference (Metal, MLX, or llama.cpp). You can choose from a selection of models based on your hardware and quality/speed preferences.

Initial release will support:

• • Local files (Markdown, plain text, PDF, common document formats)
• • Apple Calendar
• • Apple Reminders
• • Apple Notes
• • Audio/video files (transcribed locally)

Additional integrations (email, third-party apps, web bookmarks) are planned based on user demand.

When you ask a question, Nexus Note:

1. Converts your question into an embedding vector
2. Searches the vector database for semantically similar notes
3. Also performs keyword search for exact matches
4. Combines and ranks results
5. Passes the top results as context to the local language model
6. Returns an answer grounded in your actual notes, with citations

This retrieval-augmented generation (RAG) approach ensures answers are grounded in your data, not hallucinated.

Pricing hasn't been finalized. We're considering a one-time purchase model that respects the local-first philosophy—you buy the software, you own it, no subscriptions required for core functionality.

Optional cloud features (if we add them) might have separate pricing.

We're exploring options. At minimum, there will be a trial period so you can evaluate whether Nexus Note fits your workflow before purchasing.

Join the waitlist. We'll prioritize early access for users who:

• • Care deeply about privacy and local-first software
• • Have large personal knowledge stores
• • Are willing to provide detailed feedback
• • Can help stress-test the system with real workloads