On-Device AI

Meridian includes AI features that run entirely on your Mac. Content used for AI processing is not sent to a cloud AI service. If you use multi-device sync, application content is encrypted before transit.

How It Works

Meridian uses a local language model optimized for Apple Silicon. A recommended model downloads in the background, and you can continue using other features while it downloads.

What the AI Does

Action Item Extraction

Meridian identifies actionable language in messages. When someone asks you to send a file, review a document, or make a decision, Meridian can propose a to-do with the relevant context attached for you to review.

You don’t need to read every email carefully to make sure you didn’t miss a commitment. Meridian catches it for you.

Activity Summarization

Get a quick 2-3 sentence overview of your current workload:

• What needs your attention
• What’s in progress
• What’s been completed recently

This is your daily briefing — a fast way to orient yourself without scrolling through lists.

Natural Language Task Creation

Describe what you need to do in plain language and Meridian will parse it into a structured task with the right title, due date, priority, and type. Instead of filling out a form, just type what you’re thinking.

Requirements

• Apple Silicon (M1, M2, M3, or newer) is required
• Available models range from roughly 270 MB to 4.4 GB; the recommended choice depends on your Mac
• Meridian detects your GPU core count and Neural Engine to optimize performance

Privacy

This is the most important part: the AI runs on your device and nowhere else. Your emails, tasks, and calendar data are never sent to a remote server for AI processing. There is no cloud AI. Your data is not used to train models. The model runs locally using your Mac’s hardware, and that’s it.