Your meeting can stay where it happened

You have ten minutes before a sensitive client call. The Zoom invitation is open, your headphones are on the desk and you are comparing two apps before deciding which one gets to hear the conversation.

One promises local notes. Another promises bot-free recording. Neither phrase reveals whether the audio travels to a server before the finished note comes back to your computer.

This is where an easy assumption can send the meeting somewhere you did not expect. Stored locally and processed locally are two separate tests. The first locates the artifacts after the call.

The second follows the audio while the app turns it into a transcript, separates the voices and produces the note. We designed Noats to pass both tests by default.

Noats keeps its own path on your Mac

By default, recording, transcription, speaker separation and the written-up note all happen on your Mac. There is no Noats server that receives your meetings. That processing path matters as much as the location of the finished files. Your audio, transcripts, speaker labels, notes, folders and chat history are written to a database and plain files in your own Application Support folder.

Any note exports as markdown with a keystroke. These are files you own, rather than entries that can be reached only through an account on somebody else’s service. A local file is the receipt. The route the meeting took is the privacy boundary.

Storage answers where the artifacts rest. Processing answers where the conversation goes while those artifacts are being made. A product can save a final note on a Mac after sending live audio elsewhere.

That is why we describe the complete route through capture, transcription, speaker separation, note generation and storage. Local-first architecture is concrete only when every stage has an address.

The app contains no analytics and no telemetry. Automatic updates can be switched off. Our detailed map of Noats network activity follows each connection and explains what changes when you select a connected provider.

The work stays beside you

Transcription and speaker separation run on Apple silicon, so once the models are on your Mac a meeting on a plane with the network off is transcribed the same as one at your desk. Noats requires macOS 14.2 or later on Apple silicon. This is the Noats path. Zoom, Meet, Teams and the rest still carry the call itself under their own terms.

Noats records the microphone and the system audio from the Mac. It notices when an app takes the microphone and offers to record. Recording begins when you start the note, with nothing to connect first. No additional participant appears in the call.

Speaker separation also happens within that meeting. Noats treats the microphone channel as your side of the call. Voices on the other end of the call are split into speakers. Noats lets you name each one during the call or after it, and those names belong to that meeting.

We explain that per-meeting approach in how Noats tells speakers apart. There is no enrolment step. You can watch the transcript develop, apply names when you know them and keep the resulting labels with the meeting files. The architecture follows the actual call: one microphone side associated with you, remote voices separated from the system audio and a note written from the resulting transcript on the same Mac.

Follow the meeting from capture to export

Precise wording makes comparisons more useful. Meetily documents local recording, transcription and storage, with local or external providers available for summaries. Its architecture description identifies a local database for meeting metadata, transcripts and summaries alongside local transcription. Those details let a reader inspect storage and processing separately.

Mumble describes a Local Mode that keeps recordings, transcripts and notes on the Mac while performing transcription, speaker labeling and summarization there. It also describes a separate cloud mode. Its August 2026 comparison said the Mac app was being redesigned and was not then available to download. Mode names and current availability both belong in a practical comparison.

Other architectures draw the boundary elsewhere. Otter states that it uses AWS S3 for data storage. Notion sends live meeting audio to subprocessors for transcription, then deletes its temporary local audio copy after successful processing or within 24 hours. Failed processing can cause an upload to Notion that is retained for up to three days.

Noats also gives you an explicit cloud-model boundary. By default your notes are written up by a model running on your Mac. You can connect Anthropic, OpenAI or OpenRouter with your own key, point Noats at an OpenAI-compatible server or use your ChatGPT plan through your existing Codex CLI session.

Only the content needed for that request goes directly to the selected provider. Nothing routes through us.

API keys you enter are encrypted with macOS's own secure storage, backed by your Keychain.

Before you press record on your next sensitive call, follow the audio all the way through capture, transcription, speaker separation, note generation and storage. We think every local claim should survive that practical test. With Noats, the path finishes beside you. After the call, the conversation remains a set of files on your Mac, with the note exportable as Markdown in one keystroke.