Daisy

A local-first meeting recorder, push-to-talk dictation tool, and AI-notes app for macOS — with a local MCP server so Claude Desktop and Cursor can query your transcripts without anything leaving the Mac.

Daisy captures meeting audio (microphone + system-audio loopback via ScreenCaptureKit), transcribes it on-device with Whisper on the Neural Engine, and produces a Granola-style outline with action items and a draft follow-up. Audio and transcripts never leave the Mac unless you explicitly enable a cloud LLM provider for the summary step — and even then you supply your own API key.

End-user installation, FAQ, and the privacy story live at https://mydaisy.io. This README is for people building Daisy from source.

What it does

Three capture modes, one app:

• Meetings — records both sides of a call (your mic + the other side via system-audio loopback), no bot joining the meeting. On-device transcription + diarization (Remote A / Remote B, with optional mic-side attribution), a summary, action items, and a draft follow-up.
• Push-to-talk dictation — hold a hotkey, speak, and the text is pasted at your cursor in any app. Whisper by default, or on-device Parakeet (FluidAudio) for lower latency; a custom-vocabulary dictionary fixes names/jargon, and a rolling 24-hour history lets you re-copy.
• Voice notes — quick one-off thoughts saved to your Library. Optional: import existing Apple Voice Memos as flat transcripts (on-device, opt-in, needs Full Disk Access).

The differentiator: Daisy ships a local MCP server bound to 127.0.0.1 that exposes your sessions as a queryable, actionable data source to any MCP client (Claude Desktop, Cursor). Because the transcript is already local, Daisy can be a local-only MCP source — something cloud meeting tools structurally can't offer.

Status

• Latest release: see scripts/release-notes/ and https://mydaisy.io/appcast.xml. Beta ships from main; stable is promoted from a soaked beta (see RELEASING.md).
• Deployment target: macOS 14 Sonoma. The Apple Intelligence summarizer requires macOS 26 Tahoe; everything else runs on 14+.
• Apple Silicon (M1+). Signed with Developer ID, notarized, stapled, Sparkle EdDSA-signed for in-app updates.
• License: Apache 2.0 (see LICENSE). Full public source — build it and verify there's no telemetry.

Build from source

Requirements:

• Xcode 16+ with the macOS 26 SDK installed
• An active Apple Developer account if you want a signed local build (unsigned builds are fine for development inside Xcode)

Clone and open:

git clone https://github.com/addicted-studio/daisy-app.git
cd daisy-app
open Daisy.xcodeproj

The Swift Package Manager dependencies (Sparkle, WhisperKit/ArgmaxCore, FluidAudio) resolve on first project load. Hit Run; the app launches.

Project layout

Daisy/                  → SwiftUI app sources (PBXFileSystemSynchronizedRootGroup)
DaisyTests/             → unit tests
DaisyUITests/           → UI tests
Daisy.xcodeproj/        → Xcode project
scripts/
  release.sh            → end-to-end release: archive → notarize → DMG → sign → Sparkle appcast
  release-notes/        → per-version markdown bullets consumed by release.sh
  dmgbuild_settings.py  → dmgbuild config (Python) for the installer DMG
  assets/               → DMG background, app icons
build/                  → archive output (gitignored)
RELEASING.md            → branch/channel model and the release/promote/hotfix flows

Key services that drive the app:

• CoreAudioMicRecorder — CoreAudio mic capture with route-change recovery and the archive .caf writer (replaced the old AVAudioEngine tap to fix route-change/Bluetooth dropouts)
• SystemAudioCapture — SCStream loopback for the remote side of a meeting, Bluetooth-output detection, silent-capture warnings
• Transcriber / WhisperEngine — WhisperKit on-device transcription with a Silero V

... View full README on GitHub