Meeting Notes Workflow

Version: This guide describes Parrot Scribe 1.1.0.

Parrot Scribe captures and transcribes meetings on your Mac. Use this workflow to review speakers, find decisions later, and optionally pass transcript context to an AI assistant through MCP.

1. Capture or import the meeting

For meetings on your Mac, capture microphone audio, system audio, or both from the menu bar.

If the meeting happened elsewhere, open the Library and import the recording. You can drag in a Voice Memos .m4a file, another audio file such as MP3, WebM or Opus, or a video file such as .mp4 or .mov that has an audio track.

Reference: Recording Controls

2. Review the speakers

Open the session in the Library and choose Identify Speakers to review names. Session-only naming works with voice recognition off.

To reuse a voice in later recordings, first enable Recognize Saved Voices in Settings > Transcription and review its notice. Then explicitly choose Remember [name]'s voice during speaker identification and confirm Save Voice Profiles. The Remember box starts unchecked. Make sure you have any required permission from the people you record.

Reference: Speaker Recognition

3. Review and retrieve the transcript

Use the Library to review the transcript and return to earlier conversations. Search the archive when you need the wording of a decision, constraint, or commitment.

The transcript remains the source record. Notes, summaries, and follow-up drafts can be created from it when needed.

Reference: Library

4. Create notes with an AI assistant

Parrot Scribe Pro includes optional AI access through MCP. In Settings > AI Apps, choose Connect for Claude, Codex or Cursor and follow that app's confirmation. For another compatible client, create a separate token under Advanced. Connect each app separately. Simple access lets the app read the library, control recording and name speakers; Trusted access also permits the destructive operations described in MCP Server.

Use the assistant to draft meeting summaries, action items, or follow-up material. Review the result against the transcript before using it as a record of the meeting.

The assistant controls what happens after it receives transcript data, including whether it sends that data to an external service. Check the assistant's privacy settings before granting access.

Reference: MCP Server

5. Use a local model when needed

To keep the AI step local, use an MCP-compatible host configured with a local model runtime on capable hardware. Parrot Scribe provides the transcript and MCP access. The host, model, and runtime remain separate choices.

The workflow stays the same:

1. Capture or import the meeting.
2. Review speakers and the transcript.
3. Ask the connected client to retrieve the transcript through MCP.
4. Ask the local model for the notes or summary you need.

Where Whisper fits

Parrot Scribe 1.1.0 uses NVIDIA Nemotron speech models on your Mac. It does not use a Whisper runtime. The page URL is retained for people comparing a self-built Whisper workflow with a ready-made Mac app.

For a comparison with building a Whisper-based meeting-notes stack yourself, read Whisper for Local Meeting Notes on Mac.