Privacy Policy

Last updated: August 28, 2026

Muesli is local-first. By default, speech-to-text processing runs on your Mac and audio never leaves your device. If you explicitly select an optional hosted dictation provider, Muesli sends that dictation audio to OpenAI or through OpenRouter to your selected upstream model for transcription. We do not operate servers that receive, store, or process your data.

What Muesli Does

Muesli is a macOS application for dictation and meeting transcription. It uses on-device machine learning models (CoreML / Apple Neural Engine) to convert speech to text entirely on your Mac by default. Optional OpenAI and OpenRouter dictation providers are available only when you choose and configure them.

Data That Stays on Your Device

The following data is created and stored locally on your Mac. It is not transmitted to any external server unless you explicitly enable an optional cloud service described below:

Audio recordings — microphone and system audio captured during dictation and meetings; dictation audio is sent to a hosted service only while you have explicitly selected OpenAI or OpenRouter as the dictation provider
Transcripts and meeting notes — stored in a local SQLite database
Configuration and preferences — stored in a local JSON file
Speech-to-text models — downloaded once from public model repositories (HuggingFace) and cached locally

Optional Cloud Services

Muesli offers optional integrations that connect to external services. These are opt-in and require explicit user action:

OpenAI dictation (OpenAI API) — if you select OpenAI as your dictation provider and add your own API key, microphone audio is streamed directly to OpenAI's Realtime transcription API. OpenAI's API data controls explain how API data is used, retained, and excluded from model training by default; see OpenAI's API data-controls documentation for the current retention and data-sharing settings.
OpenRouter dictation — if you select OpenRouter as your dictation provider, connect an account or key, and choose a transcription model, the completed dictation recording is sent through OpenRouter to that selected upstream model. Processing is governed by OpenRouter's and the selected model provider's policies.
Meeting summarization (OpenAI API / OpenRouter) — if configured, meeting transcripts are sent to the selected API provider to generate summaries. This is governed by the respective provider's privacy policy.
Meeting summarization (ChatGPT subscription) — if you sign in with your ChatGPT account, transcripts are sent to OpenAI's API for summarization using your existing subscription.
Sparkle update checks — Muesli checks for new versions via an appcast feed hosted on GitHub. No personal data is transmitted.

Google Calendar Integration

Muesli can optionally connect to your Google Calendar to display upcoming meetings in the status bar and auto-detect meeting start times. When you connect Google Calendar:

Muesli requests read-only access to your calendar events (event titles, times, and durations)
Calendar data is fetched directly from Google's API to your Mac — it does not pass through any intermediary server
Calendar data is used only for displaying upcoming meetings and is not stored persistently
OAuth tokens are stored in your macOS Keychain, protected by the operating system
You can revoke access at any time from your Google Account permissions page

macOS Permissions

Muesli requests the following macOS permissions, each for a specific purpose:

Microphone — to capture your voice for dictation and meeting transcription
Screen Recording — to capture system audio during meetings (via ScreenCaptureKit)
Accessibility — to paste transcribed text into other applications
Input Monitoring — to detect the global dictation hotkey
Calendar — to read upcoming events from your local macOS calendar (EventKit)

Anonymous Usage Analytics

Muesli uses TelemetryDeck to collect minimal, anonymized usage signals (e.g., which features are used and app launch counts). When Muesli detects a hard failure, it may also send an allowlisted diagnostic category, workflow stage, selected on-device transcription backend/model, app version and build, operating system and device model, and a random incident ID. These diagnostics never include raw error messages, audio, transcripts, meeting or calendar titles, clipboard or screen contents, API keys, auth tokens, local file paths, raw logs, database contents, or other personal content. TelemetryDeck does not store IP addresses, and identifiers used for anonymous user and session counts cannot be traced back to a person. You can learn more about TelemetryDeck's privacy practices on their privacy page.

No Personal Data Collection

Muesli does not collect, transmit, or store personal data on external servers that we operate. Apart from the anonymous usage signals described above, data remains on your device unless you choose an optional cloud provider such as OpenAI or OpenRouter dictation, hosted cleanup or meeting summaries, or Google Calendar sync. The application remains functional without an internet connection when you use its local features.

Open Source

Muesli's source code is publicly available on GitHub under the MIT license. You can inspect exactly what the application does with your data.

Children's Privacy

Muesli is not directed at children under 13. We do not knowingly collect data from children.

Changes to This Policy

If this policy changes, the updated version will be posted here with a revised date. Muesli does not have a mechanism to notify users of policy changes since we do not collect contact information.

Contact

For questions about this privacy policy, open an issue on GitHub or reach out to the maintainer via the repository.