Oats

Oats is a free, open-source desktop app from Ariso that records your meetings directly on your Mac or Windows PC and turns them into structured notes and action items — without ever sending a bot into the call. It lives in the menu bar, stays out of the way while you meet, and hands you a clean transcript and summary when you're done. The core is licensed under MIT and the full source is public on GitHub.

What It Is

Oats is a local-first meeting notes tool built with Tauri v2, Vue 3, and Rust. It captures system audio directly — no browser extension, no third-party recording bot — and offers two transcription backends: a free cloud path through Ariso, and a fully offline path that runs transcription and note generation entirely on-device using Apple Neural Engine (CoreML/MLX on Apple Silicon) or Parakeet ONNX on Windows. Notes are stored as plain Markdown, and action items can sync to an Obsidian vault. The project is positioned as a privacy-first alternative to cloud-only meeting recorders.

How the Two Backends Work

Oats lets users choose between two modes at any time via Settings → Transcription Backend:

• Ariso (cloud): Sign in with Google or Microsoft, hit record, and audio streams to Ariso's backend for real-time transcription and note generation. No model training on user data, per Ariso's published privacy policy. Requires an account and internet connection.
• Local (on-device): Flip one switch and everything — recording, transcription, speaker labels, and AI notes — runs offline. No login, no upload. Audio, transcripts, and notes never leave the machine. A one-time model download is required (speech model + language model), after which the app works completely air-gapped.

Local files are stored under ~/.ariso/recordings/ as recording.mp3, transcript.md, and ari-note.md.

Key Capabilities

• One-click recording from the menu bar with pause/resume/stop controls
• Real-time streaming transcription as people talk
• Speaker labeling (who said what) — available on the Ariso cloud backend; on-device support noted as in progress
• Automatic Markdown meeting summary with action items per person and a meeting quality score
• Searchable library of all past recordings, transcripts, and notes
• Todos tab that collects action items across meetings; checking one off closes it in the notes
• Obsidian vault integration: notes stored as plain Markdown, action items as Obsidian Tasks
• Export and download of recordings and notes in one click
• Sign in with Google or Microsoft
• Works with Zoom, Google Meet, Teams, phone calls, or any audio the Mac can hear

Platform Support

Oats targets macOS 14+ on Apple Silicon as its primary supported platform, with full Ariso and local backend support. Windows 11 is available as a release preview — signed x64 NSIS installer and in-app auto-updating work, but system-audio capture, auto-record detection, and native sharing are still being finalized for broad public support. Windows 10 is planned for smoke coverage but is not yet public-support ready. The app is code-signed and notarized by Apple and supports Homebrew installation via a dedicated tap.

Update: v0.24.0

The latest release is v0.24.0, published September 13, 2026. Recent releases visible on the product page include:

• v0.23: Todos tab for cross-meeting action item tracking; Obsidian vault integration for local backend; Microsoft sign-in support
• v0.21: Download and export of recordings and notes
• Speaker tagging (Ariso backend): voice-to-person matching so notes attribute speech by name rather than "Speaker 2"

The GitHub repository shows active weekly shipping cadence, with the last push on September 14, 2026. The project had 43 stars and 5 forks as of the latest metadata snapshot.

Open-Source Architecture

Oats is built on Tauri v2 (Rust backend, Vue 3 frontend), with on-device inference powered by FluidAudio, MLX Swift LM (Apple Neural Engine), and Parakeet ONNX. The full source is public at github.com/ariso-ai/oats under the MIT License. Ariso publishes the complete changelog on GitHub Releases and welcomes pull requests via a documented CONTRIBUTING.md. The local-backend design means the app can be self-hosted or forked without any dependency on Ariso's cloud infrastructure.