Quartz

About Quartz

Quartz is an AI-native email client for macOS that runs locally on the user's machine. It sorts Gmail messages by importance, learns individual writing style over time, and generates reply drafts. The AI engine processes data on-device using the Gemma 4 E4B model, so emails remain end-to-end encrypted and are not sent to third-party AI providers.

Review

Quartz enters the email client space with a local-first architecture that keeps processing on the Mac rather than routing data through cloud servers. The application is currently in public beta and free to use. It filters incoming mail into five standard importance levels and adapts its behavior based on user corrections and sent email patterns.

Key Features

• On-device AI processing with Gemma 4 E4B, keeping mail end-to-end encrypted and off AI provider servers.
• Importance-based sorting that places every incoming message into one of five standard levels; users can correct classifications and the system incorporates that feedback for similar future emails.
• Draft generation that pulls examples from emails sent within the same thread first, then messages to the same contact, then others. A writing style profile in settings lets users supply additional context immediately.
• Feedback loop for filtering personalization: when a user re-categorizes an email and shares feedback, the model loads that feedback when encountering similar senders or content.
• Built with Tauri, which avoids bundling a full Chromium runtime and keeps the native Mac app compact.

Pricing and Value

The tool is currently in public beta and free to use. The makers have not published a pricing model for the post-beta period, so long-term cost structure is not yet defined.

Pros

• Processing stays on the user's own Mac; no email data leaves the device for AI training or inference.
• The model update approach uses a CLAUDE.md-style profile to bridge the cold-start gap before enough sent mail accumulates.
• Writing style adaptation draws from a hierarchy of examples-thread, contact, then general history-which can become noticeable after a few replies.
• Filtering learns from direct user corrections rather than relying solely on pre-set rules.
• The Tauri-based build reduces application footprint compared to Electron-based alternatives.

Cons

• Model updates require re-downloading the full 4GB Gemma 4 weight file; no differential update or LoRA adapter mechanism is in place yet.
• Custom user-defined email categories are on the roadmap but not available in the current beta, limiting organizational flexibility to the five built-in levels.
• The tool is not well suited for users who rely heavily on cross-platform workflows or non-macOS devices, since it runs only on Mac.

Quartz fits professionals who handle a high volume of Gmail on a Mac and want AI-assisted triage without sending their inbox to a cloud service. The local-only architecture and on-device model make it relevant for anyone handling sensitive or regulated correspondence, though the lack of custom categories and cross-platform support may rule it out for teams with diverse device setups.

Open 'Quartz' Website