LogoTopAIHubs
icon of Daisy

Daisy

Local-first, open-source Mac meeting recorder and dictation app.

Introduction

What is Daisy

Daisy is a free, open-source, local-first meeting recorder and dictation app for Mac (Apple Silicon, macOS 14+). It records meetings from any app (Zoom, Meet, Telegram, etc.) by capturing both your microphone and the remote audio, transcribes on-device using WhisperKit on the Neural Engine, and saves transcripts as Markdown files to a folder of your choice (e.g., Obsidian vault, iCloud Drive). Daisy also includes a push-to-talk dictation mode and a local MCP server that lets Claude Desktop, Cursor, or any MCP client query your transcripts directly — all without data leaving your machine unless you explicitly send it to a destination like Notion, Linear, Attio, Slack, or a webhook.

How to use Daisy
  1. Download and install the DMG from the website (Apple Silicon M1+, macOS 14+).
  2. Launch Daisy — a small flower-shaped widget appears at the edge of your screen.
  3. Start a meeting recording by clicking the widget or using a keyboard shortcut. The centre turns orange and petals dance with audio.
  4. Stop recording — Daisy automatically transcribes the audio on-device and saves a Markdown file to your chosen folder.
  5. Review transcripts in your folder or via the Daisy Library.
  6. Optional: Enable the MCP server in Settings to let Claude Desktop or Cursor read and act on your transcripts.
  7. Optional: Set up destinations (Notion, Linear, Attio, Slack, webhook) to automatically push transcripts after recording.
Features of Daisy
  • Local recording — captures both your mic and remote audio from any app; no bot joins the meeting.
  • On-device transcription — WhisperKit on the Neural Engine, with Silero VAD pre-pass to avoid hallucinating silences.
  • Speaker diarisation — offline re-diarisation using FluidAudio (pyannote family) for ~10–15% diarization error; two-channel capture (you vs. remote) reduces confusion.
  • Markdown transcripts — saved to any folder you choose; plain files you own.
  • MCP server — bound to 127.0.0.1, exposes tools like list_sessions, get_session, search_sessions, rename_speaker, route_session_to_destination.
  • Push-to-talk dictation — lilac centre, audio-to-text anywhere you type.
  • Voice notes — quick one-off thoughts saved to your Library.
  • Destinations — send transcripts to Notion, Linear, Attio, Slack, or any webhook; folder-based auto-routing.
  • BYOK AI — use your own Anthropic, OpenAI, Apple Intelligence, Ollama, or LM Studio key for summaries; Daisy never sees your key or data.
  • No telemetry, no account, no subscription — free during beta, lifetime after launch.
Use Cases of Daisy
  • Meeting notes for remote teams — record Zoom/Meet calls, get accurate transcripts with speaker labels, and push action items to Linear or Notion.
  • Personal journaling or dictation — use push-to-talk mode to capture thoughts hands-free.
  • Compliance-sensitive environments — transcripts stay on your Mac; no cloud server to audit.
  • AI-powered workflow — connect Daisy’s MCP server to Claude Desktop to query past meetings, rename speakers, or re-summarize.
  • Obsidian integration — save transcripts directly into your Obsidian vault for linking and searching.
Pricing

Daisy is free during beta and will have a lifetime license after launch — never a per-meeting subscription.

FAQ

Q: Does Daisy upload my recordings anywhere? A: No. Transcription runs entirely on your Mac. Transcripts are saved to your chosen folder. Nothing leaves your machine unless you explicitly send a transcript to a destination (Notion, Linear, etc.) or use a cloud AI key for summarization.

Q: What hardware do I need? A: Apple Silicon (M1 or later) and macOS 14+.

Q: Can I use my own AI model for summaries? A: Yes. Daisy supports BYOK for Anthropic, OpenAI, Apple Intelligence, Ollama, LM Studio, or any MCP-compatible local model.

Q: How accurate is the transcription? A: On-device WhisperKit achieves ~95% word accuracy. Speaker diarization error is ~10–15% offline (on par with pyannote). Two-channel capture (you vs. remote) improves accuracy.

Q: Can I rename speakers? A: Yes. Rename a speaker once, and Daisy remembers their voice for future recordings.

Q: Is there a cloud component? A: No. Daisy has no cloud server, no telemetry, no account system. The MCP server listens only on 127.0.0.1.

Q: How do I get transcripts into Claude Desktop? A: Enable the MCP server in Daisy’s Settings, click “Add to Claude Desktop”, and restart Claude. Your transcripts become a queryable data source.

Q: Can I send transcripts to other tools? A: Yes. Daisy supports Notion, Linear, Attio, Slack (webhook), and generic webhooks. Configure destinations in Connections.

Q: Is Daisy open source? A: Yes, under Apache 2.0. Source is on GitHub.

Newsletter

Join the Community

Subscribe to our newsletter for the latest news and updates