Reyn — Local AI for Mac That Watches Your Screen

TL;DR

TL;DR: Reyn is a Mac app that runs a local AI on Apple Silicon, continuously watches your screen to build a private work journal, and answers questions about your past activity — all without sending any data to the cloud.

What Is Reyn?

Reyn is a native macOS app that treats your screen as a live context feed for a personal AI companion. Unlike cloud-based AI assistants that process everything on remote servers, Reyn runs entirely on your Mac’s Apple Silicon Neural Engine (ANE). It captures your screen continuously, builds a searchable journal of your work activity, and lets you query that journal through a conversational interface — all without a single byte leaving your machine.

“Privacy-first desktop AI that watches your screen, builds a private journal of your work, and answers questions about your week — without sending your data to the cloud.”

Source and Accuracy Notes

• Project page: usereyn.com
• Source repository: No public GitHub repo found (native macOS app, closed source)
• License: Proprietary (pre-order model)
• HN launch thread: news.ycombinator.com/item?id=47334565 (via AgentMBOX mention)
• Source last checked: 2026-06-26

How Reyn Works

Screen Capture Architecture

Reyn uses macOS screen capture APIs to sample your display at configurable intervals. The captured frames are processed locally through a vision model running on the ANE — Apple’s dedicated neural processing hardware on M-series chips. This means screen analysis happens on-die, without invoking any cloud service.

The journal stores structured metadata (not raw screenshots) — timestamps, app context, and extracted activity summaries. Screenshots themselves are held in a local encrypted cache and never transmitted.

Local AI Inference

The companion model runs via Core ML on Apple Silicon. Reyn supports integration with external AI providers as an optional cloud fallback, but the default and privacy-preserving mode is fully local.

Privacy Model

Reyn’s core design principle is local-first. Your work journal, screen activity, and query history never leave your Mac unless you explicitly enable an opt-in cloud sync feature. Even then, data is encrypted end-to-end.

Practical Evaluation Checklist

Setup:

• [ ] Download the pre-order build from usereyn.com
• [ ] Requires macOS on Apple Silicon (M1 or newer)
• [ ] Grants screen capture permission on first launch
• [ ] Configure capture interval (default: every 30 seconds)
• [ ] Pin the app to menu bar for quick access

Core features to try:

• [ ] Ask “what was I working on Tuesday afternoon?”
• [ ] Search your journal by app, keyword, or time range
• [ ] Enable the floating companion widget
• [ ] Configure heartbeats for passive journaling during deep work

Privacy controls:

• [ ] Review what data is stored in local cache
• [ ] Disable cloud sync if enabled (opt-in)
• [ ] Set app-level screen capture schedules

Security Notes

• All screen analysis runs on-device via Apple Neural Engine
• Journal data stored in local encrypted SQLite database
• No telemetry, no outbound connections by default
• App Sandbox restricts access to system resources
• Memory-mapped screen buffers are zeroed after each capture cycle

Reyn vs. Alternatives

| Feature | Reyn | Recall.ai | Obsidian + AI | Logseq | |---|---|---|---|---| | Always-on screen capture | Yes | Yes | No | No | | Fully local processing | Yes | No | Yes (with local LLM) | Yes | | Built-in work journal | Yes | No | Manual | Manual | | macOS native | Yes | Browser-based | Cross-platform | Cross-platform | | Cloud optional | Yes | Required | No | No |

FAQ

Q: Does Reyn always watch my screen? A: Reyn captures your screen at intervals you configure (default every 30 seconds). You can pause or disable capture at any time. Captured frames are processed locally and either discarded or stored in your local journal depending on your settings.

Q: Is my screen data sent to the cloud? A: No — by default, Reyn stores everything locally. There is an optional iCloud sync feature for cross-device access, but it uses end-to-end encryption. No data is sent to Reyn’s own servers.

Q: What Apple Silicon chips are supported? A: Reyn requires M1, M2, M3, or M4 series chips with the Neural Engine. The app is optimized for the Unified Memory architecture of Apple Silicon.

Q: How does Reyn compare to cloud AI assistants? A: Cloud assistants like ChatGPT or Claude offer broader general knowledge but require sending your prompts to external servers. Reyn’s knowledge is limited to your own screen activity and work history — making it more of a personal work memory than a general AI assistant.

Q: What is the pricing model? A: Reyn is currently available via pre-order. Exact pricing details are on the usereyn.com website.

Conclusion

Reyn occupies a distinct niche: a privacy-first, local AI companion that turns your daily screen activity into a searchable work journal. For users who want AI assistance grounded in their own context — without surrendering that context to a third-party cloud — Reyn’s on-device approach is compelling. The always-on screen capture differentiates it from passive journaling tools, and the local-only processing addresses the privacy concern that typically makes people hesitant to adopt such tools. If you’re on Apple Silicon and want your AI to actually know your work, Reyn is worth a pre-order.