On-Device AI

Private, local AI capabilities powered by Apple Silicon

Meridian includes built-in AI that runs entirely on your Mac. It can identify action items in messages, summarize context, help structure tasks, draft replies, and improve search without sending your content to a cloud AI service.

How It Works

Meridian runs a language model locally on your Mac using Apple Silicon’s Metal GPU acceleration. All AI processing happens on your device. No data is sent to any cloud AI service.

Requirements

Meridian requires an Apple Silicon Mac (M1 or later). It downloads a recommended model in the background; available models range from about 270 MB to 4.4 GB, and the default choice depends on your hardware. The rest of Meridian remains usable while a model downloads.

Action Item Extraction

The AI identifies actionable language in messages and proposes structured to-dos with useful context. You review the suggestion before adding it to your list.

Activity Summarization

Get concise summaries of message threads and activity context so you can catch up without rereading every item.

Natural Language Tasks

Describe a task in plain English and Meridian turns supported details — such as a title, due date, priority, recurrence, or location — into structured fields for you to review.

Smart Search

Alongside keyword matching, Meridian uses a local semantic index to find related content by meaning. Embeddings stay in a per-device index and are not synced.

Privacy Guarantee

The AI model runs entirely on your device. Your data never leaves your Mac for AI processing. No cloud AI APIs are used. The AI model is not trained on your data. This is a core architectural decision, not a setting you need to enable.