Open-source project
digimata/quill avatar
digimata/quill

quill: the macOS meeting recorder where nothing leaves the machine

Ultra-minimalist macOS recording + transcription.

3,927 stars283 forksSwiftMIT

At a glance

What is it?
quill is digimata's MIT-licensed Swift application for fully local meeting recording and transcription on macOS 15 and later, capturing microphone and call audio as separate tracks, transcribing on-device, and writing a timestamped me and them transcript into ~/Recordings. Windows and Linux ports are planned as native independent implementations sharing only session layout and transcript format, and it is the sibling of the parrot project.
Who is it for?
Use quill when meeting transcription must stay on the machine, for privacy, compliance or offline use, and your Mac runs macOS 15 or later, with Apple Silicon recommended for transcription speed. It is macOS-only today, the Windows and Linux columns are a capture probe planned and proposed, so verify platform status before planning around it.
Can I use it commercially?
Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
Is it still maintained?
Yes. The repository last received commits 2 days ago.
What is it written in?
Mainly Swift, according to GitHub's language statistics.

Answers come from the project's GitHub data, last synced on September 30, 2026, and from our analysis. They are not legal advice.

Editorial analysis

Two tracks, one transcript

quill records your microphone and the call's audio as separate tracks, transcribes them on-device, and writes a timestamped transcript marked me and them, with the privacy statement attached to the design, nothing leaves the machine. The separate-track choice is what makes the speaker attribution work without speaker diarization models, since the me track and the them track are known by capture source rather than inferred from mixed audio, and the on-device transcription means the speech model runs on the Mac's own silicon, the reason Apple Silicon is recommended for transcription speed. The output is Markdown, transcript.md beside the audio in a timestamped session folder, immediately readable and greppable without any app.

Three commands to installed and running

The install path is a build script, an install script and an optional launch registration:

sh
./scripts/build-macos
sudo ./scripts/install-macos
quill install --launch-at-login   # optional

with the requirement of macOS 15 or later. Usage is three steps, click the feather in the menu bar and choose Start recording, where first use asks for microphone and system audio permissions, choose Stop recording when the meeting ends and transcription starts automatically, and find the transcript at ~/Recordings with a yyyy.MM.dd-HHmm folder name beside the audio. The menu bar presence means the tool is a background utility rather than a windowed app, and launch at login is opt-in rather than default.

Native per platform, shared by convention

The platforms table states status precisely, macOS 15 and later working, Windows 11 with a capture probe planned, Linux with a capture probe proposed, and the architecture note beneath carries the design decision, each platform is a native, independent implementation sharing the session layout, transcript format and test fixtures, not source code. The repository's top level mirrors this, macos, windows and linux directories each holding their own implementation and README, with docs/architecture.md documenting the shared contracts. For a project whose core value is on-device machine learning, going native per platform trades code reuse for using each system's own speech and audio stacks, and the shared fixtures keep the outputs comparable across those stacks.

The privacy posture, stated in one line

Nothing leaves the machine is the entire privacy policy, and it is a functional property rather than a policy page, since transcription happens on-device there is no upload step to audit, no API key to leak, and no service whose terms could change. Recording both sides of the call locally also carries responsibilities the user owns, consent from the other participants being a legal matter in many jurisdictions, and the tool's design, separate tracks and local storage, leaves that governance in the user's hands rather than mediating it. The session folder layout, timestamped directories holding audio plus transcript, makes retention management a filesystem operation, delete the folder and the meeting is gone.

Swift, menu bar shaped, and unreleased

The project is written in Swift, MIT licensed, and hosted on the master branch with the repository pushed 2026-09-27, three days before this writing. There are no GitHub releases, so the build-and-install scripts are the distribution, a source-first model matching the tool's privacy posture, since building locally means the binary that runs is the one you compiled. A CHANGELOG.md sits at the root tracking changes between builds, and the sibling project parrot under the humanitas-labs organization is named as a relative, the two tools sharing an authorial lineage if not code.

What the name borrows

The quill name reaches for the writing instrument rather than the many unrelated products sharing it, office supplies, a mall, a bike stem, a video game item, and the README invests nothing in disambiguating them, its identity carried by the one-line pitch instead. Within the tool itself the name surfaces literally, the feather in the menu bar is the interface, a recording indicator dressed as the writing tool it replaces. For a search-driven evaluation, the practical handle is the combination of terms, quill meeting recorder macOS, since the single word alone lands elsewhere entirely.

Editorial conclusion

Use quill when meeting transcription must stay on the machine, for privacy, compliance or offline use, and your Mac runs macOS 15 or later, with Apple Silicon recommended for transcription speed. It is macOS-only today, the Windows and Linux columns are a capture probe planned and proposed, so verify platform status before planning around it. Before installing, expect the two permission prompts for microphone and system audio on first recording, note that the install script uses sudo to place the binary, and read the macOS documentation for configuration, the CLI and troubleshooting rather than guessing at options.

Frequently asked questions

What is the quill macOS app?

quill is a fully local meeting recorder and transcriber for macOS 15 and later, MIT licensed and written in Swift. It records the microphone and the call's audio as separate tracks, transcribes on-device, and writes a timestamped me and them transcript as Markdown in ~/Recordings, with nothing leaving the machine.

how to install quill?

Build and install with ./scripts/build-macos followed by sudo ./scripts/install-macos, and optionally run quill install --launch-at-login for login startup. macOS 15 or later is required, and Apple Silicon is recommended for transcription speed. First recording prompts for microphone and system audio permissions.

Does quill work on Windows or Linux?

Not yet. The platforms table marks macOS 15 and later as working, Windows 11 with a capture probe planned, and Linux with a capture probe proposed. Each platform is implemented natively and independently, sharing the session layout, transcript format and test fixtures rather than source code.

Official sources

  1. digimata/quill on GitHub
  2. Issues
  3. License: MIT
  4. README
Add this badge to your README

If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.

Add this badge to your README

markdown
[![Hysen Labs](https://hysenlabs.com/badge/digimata-quill.svg)](https://hysenlabs.com/projects/digimata-quill)