We open-sourced Logue — a privacy-first macOS meeting-notes + writing app that runs on-device (MLX, Apple Silicon) entirely

At Bitwize, we've been building Logue, a native macOS app for AI meeting notes and writing, and we just open-sourced it (MIT). We're sharing it here because the whole point is that it runs 100% on-device — we wanted something that could transcribe and summarize meetings without shipping audio or notes to anyone's cloud.

By default, nothing leaves your Mac. The only network calls are the initial on-device model download, app update checks, and opt-in features you explicitly turn on (web search or plugging in an external AI provider if you want one). No accounts, no telemetry, no backend.

What it does:

• Real-time transcription of mic and system audio (Apple's on-deviceSpeechTranscriber)
• Speaker diarization — who said what — via FluidAudio (streaming Sortformer)
• "Smart Minutes": local LLM summaries, action items, highlights
• Writing assistant: 60+ modes (rewrite, grammar, clarity, tone), a document editor with AI chat, vocabulary suggestions
• On-device PII detection and a fact-check/verify panel
• Templates, Spaces, and "Ask Logue" chat over your own notes

Stack: Swift + SwiftUI/AppKit, MLX (mlx-swift-lm) for LLM inference, Apple's Speech framework, FluidAudio for diarization, Sparkle for updates. Data is AES-256-GCM encrypted at rest.

Honest caveats: it targets macOS 26 (Tahoe) and Apple Silicon only (MLX + the new Speech APIs), so it won't run on Intel or older macOS. It's early — expect rough edges — and we'd genuinely love feedback, issues, and PRs.

Repo: https://github.com/bitwize-ai/Logue

Happy to answer anything about the on-device pipeline, MLX inference, or diarization in the comments — we're the team that built it.