Meetify

A local-first AI meeting copilot for macOS and Windows.
Think Cluely, but open-source, self-hosted, and free if you want it to be.

Meetify listens to your meeting (system audio + mic), transcribes it live, and gives you real-time talking points, answers, and recaps in a small floating widget you can tuck to the edge of your screen. No backend, no account, no telemetry — it runs entirely on your machine, and the only network calls it makes are the ones you explicitly choose (an LLM provider, optionally a hosted transcription API).

How it works

1. The Tauri desktop shell starts the floating widget window and applies the platform's official capture-protection setting. On Windows this uses WDA_EXCLUDEFROMCAPTURE; on macOS it uses AppKit's NSWindow.sharingType = .none. The widget stays visible to you but is omitted from supported screenshots and screen-share captures. Linux has no portable equivalent and reports that limitation.
2. Meetify captures microphone audio and system meeting audio locally, mixes the streams, and sends the audio to the transcription worker. The default Whisper path runs locally through whisper.cpp, so meeting audio does not need to leave the computer.
3. The live transcript is shown in the widget and provides context for AI Assist and the command bar. AI requests go only to the provider you selected: local Ollama, Anthropic, OpenAI, or OpenRouter.
4. Notes, transcripts, and meeting recaps are stored locally and can be exported to Markdown. Credentials are stored in the operating system keychain rather than a plaintext configuration file.

Why Meetify instead of Cluely Cluely-style copilots are usually a hosted subscription with your audio and transcripts flowing through someone else's server. Meetify does the same job differently:

Meetify Cluely-style tools Cost Free with Ollama, or pennies per meeting with your own API key Monthly subscription Where it runs 100% local — Tauri desktop app Cloud-hosted Your data Never leaves your machine unless you choose an API provider Sent to their servers Transcription Local Whisper (whisper.cpp), fully offline Cloud only Account required No Yes Source Open source (MIT) Closed Run it against a local Ollama model and it costs $0 forever. Point it at Claude, GPT, or any OpenRouter model instead and you're typically spending cents per meeting — a fraction of what a subscription copilot charges for the same real-time assist.

Features Live transcription — system audio and mic are captured and mixed, then transcribed in real time with local Whisper. Nothing about what's said needs to leave your machine. AI Assist, on demand — hit a hotkey to get talking points and suggested answers pulled from the live conversation, streamed into a compact suggestions pane. Ask AI anything mid-meeting — a command bar in the widget lets you ask follow-up questions about what's being discussed without breaking your flow. Bring your own model — plug in Anthropic, OpenAI, OpenRouter, or point Meetify at a local Ollama server. Switch providers any time in Settings. Free-tier friendly — OpenRouter's :free models and local Ollama models mean you can run Meetify at zero API cost if you want. Floating, tuck-away widget — an always-on-top bar that expands for the transcript and suggestions, collapses to a slim pill, and can be tucked to the edge of the screen when you don't need it. Meeting recaps — notes and transcripts are saved locally (SQLite) and exportable to Markdown when the meeting's done. Real keychain security — API keys are stored in the OS keychain (macOS Keychain / Windows Credential Manager), never in a plaintext config file. Zero telemetry — no analytics, no crash reporting, nothing phones home. Ever. Live AI conversation — ask questions from the chat bar, use recent transcript context, and let Meetify answer detected questions from the other participant. OpenRouter's full catalog — the model picker refreshes from OpenRouter at runtime, puts free models first, and includes every currently published model. macOS menu-bar companion — the transparent Meetify M sits beside the system status icons for one-click access to the widget and settings. How it stays private on a screen share On Windows and macOS, Meetify applies the operating system's official per-window capture-protection flag at startup (WDA_EXCLUDEFROMCAPTURE / NSWindow.sharingType = .none). The widget remains visible locally but is excluded from supported screenshots and screen-share capture APIs. Linux does not provide a portable equivalent, so sharing a specific application window remains the reliable fallback:

Zoom — Share Screen → pick the app window, not "Desktop" Google Meet — Present now → "A window" Microsoft Teams — Share → pick the window, not the whole screen The local widget still shows a small "AI Assist Active" badge (on by default), but supported capture APIs exclude the entire Meetify window from the shared image. Capture protection is best-effort: unsupported operating systems, capture tools that ignore OS-level window protection, and Linux may still show the window.

Privacy No analytics, no crash reporting, no account system. API keys live in your OS keychain and are sent only to the provider you picked. Transcripts and notes are stored locally in SQLite. Audio never leaves your machine for transcription if you use the local Whisper model (default). It only goes out if you opt into a hosted transcription API. Tech stack Shell: Tauri 2 (Rust backend + system webview) Frontend: React + TypeScript + Vite Audio: system-audio loopback + microphone, mixed to one stream Transcription: local whisper.cpp by default, hosted Whisper API optional LLM: Anthropic, OpenAI, OpenRouter, or local Ollama — bring your own key, or none at all Storage: SQLite (rusqlite) for meeting data, OS keychain for secrets Getting started Requires Node 18+ and a Rust toolchain.

npm install npm run tauri dev On first launch, Meetify walks you through picking a provider (or connecting Ollama), the local screen-share explainer, and downloading the local transcription model (~140MB, one-time).

Quick installation

Download the latest release from GitHub Releases, open the installer for your platform, and launch Meetify. On macOS, the current Apple-silicon DMG is Meetify_0.1.2_aarch64.dmg. Because the macOS build is ad-hoc signed, macOS may require Control-click → Open the first time.

After launch, choose a provider and model, paste its API key, and press Verify. OpenRouter's model list includes free entries and refreshes automatically. Ollama needs no key and stays local. The Meetify M appears in the macOS menu bar; click it to bring the widget forward or use the menu to quit.

All app versions

Version	macOS Apple silicon	Release notes
0.1.2	Download DMG	Functional AI Assist
0.1.1	Release assets	Screen-share protection
0.1.0	Release assets	Initial desktop release

Windows builds use the same Tauri project and are published when a Windows runner produces the signed installer. Linux has no portable OS-level capture exclusion API.

Contributing PRs welcome. Keep capture protection scoped to Meetify-owned windows and use official platform APIs where available. Document platform limitations and preserve the visible local "AI Assist Active" indicator.

License MIT