The open-source, local-first AI meeting assistant and real-time interview copilot. Live two-party transcription, instant AI answers grounded in your own documents, and an overlay only you can see.
Download · Website · Docs · Contributing · Report a bug
The NexQ overlay during a live interview: real-time transcription and AI suggestions.
- Local-first and private. Audio is captured on your machine, and you choose where it goes: fully offline engines (Whisper, Parakeet, Ollama) or the cloud providers you configure. No bot joins your call.
- Free and open source. MIT licensed, no account, no subscription, no usage caps.
- Bring your own models. 11 speech-to-text engines and 9 LLM providers, from local Whisper and Ollama to Deepgram, OpenAI, Gemini, or your existing Claude Pro/Max subscription.
- Built for the moment. Answers stream in while the other person is still talking, grounded in your resume, notes, or slides.
v2.25
- macOS support (Apple Silicon, beta). Native macOS build, Keychain credential storage and the transparent overlay. A community contribution by @visitek.
- Stealth Mode: "Hide from screen share". NexQ stays visible to you but is left out of screen sharing, recordings and screenshots (
WDA_EXCLUDEFROMCAPTUREon Windows,NSWindowSharingNoneon macOS). While it's on, the overlay keeps a plain arrow cursor, and its minimize button hides it to the tray (the menu bar on macOS) during a meeting instead of opening the dashboard. NexQ also leaves the taskbar on Windows, where Alt+Tab, the tray and the shortcuts still reach it, and the Dock and Cmd+Tab on macOS, where the menu-bar icon and the shortcuts still reach it. On macOS 14 and later, some recorders that use ScreenCaptureKit can still capture NexQ, so try a test share first. Toggle it in Settings, from the dashboard or overlay header, from the tray, or with Ctrl+Shift+H (Cmd+Shift+H on macOS).- Czech transcription.
Recently shipped: transcript paragraphs with separated remote speakers, a transcript scope picker and answer history (v2.24); a Claude model picker read from the CLI with per-model effort (v2.23); Claude via your own subscription, silent background updates, and "Ask a past meeting" (v2.22). See the full CHANGELOG.
|
Transcription
|
AI copilot
|
|
Privacy and local-first
|
Meeting tools
|
| Live interview | Lecture mode | Past meeting review |
|---|---|---|
More screenshots
| Dual-party transcription | Real-time AI copilot |
|---|---|
| Local RAG pipeline | Multi-language translation |
- Download the
x64-setup.exeinstaller from the latest release. - Run the installer. If SmartScreen warns about an unrecognized app, click More info → Run anyway (the installer is not code-signed; you can always build from source).
- NexQ updates itself in the background and installs on quit, never in the middle of a meeting.
- Download the
.dmgfrom the releases page and drag NexQ into Applications. - The beta build is not notarized, so Gatekeeper will block it. Clear the quarantine flag once:
xattr -cr /Applications/NexQ.app
- Grant microphone access when prompted. Please report anything that misbehaves with the
platform: macoslabel.
- Pick a speech engine in Settings → STT Providers. Web Speech works with zero setup; download a local model for offline use, or add a Deepgram / Groq key.
- Pick an LLM in Settings → LLM Providers. Ollama is detected automatically; cloud providers need an API key; Claude can use your existing subscription.
- (Optional) Load context such as your resume or notes in the Context panel.
- Start a meeting with Ctrl+M. NexQ captures system audio, so it works with Zoom, Teams, Meet, Discord, or anything else.
Guides: Getting started · Choosing providers · Audio setup · Interview copilot · Meeting detection · Keyboard shortcuts · Troubleshooting
Speech-to-text (11)
| Engine | Runs | Notes |
|---|---|---|
| Web Speech | Browser API | Zero setup; uses the WebView's built-in speech service |
| Windows Speech | Local | Windows only, offline fallback |
| Whisper.cpp | Local | Downloadable Whisper models, tiny to large-v3 |
| Parakeet TDT | Local | NVIDIA Parakeet via ONNX Runtime, high accuracy |
| ORT Streaming | Local | Streaming transducer via ONNX Runtime |
| Sherpa-ONNX | Local | Streaming transducer in a sidecar process |
| Deepgram | Cloud | Nova-3 and others, real-time streaming |
| Groq Whisper | Cloud | Fast Whisper with a free tier |
| Whisper API | Cloud | OpenAI Whisper REST API |
| Azure Speech | Cloud | Azure Cognitive Services |
| Cartesia | Cloud | ink-2 real-time streaming with built-in turn detection |
LLM (9)
| Provider | Runs | Notes |
|---|---|---|
| Ollama | Local | Auto-detected at localhost:11434 |
| LM Studio | Local | OpenAI-compatible local server |
| OpenAI | Cloud | API key |
| Anthropic | Cloud | API key |
| Google Gemini | Cloud | API key; optional Gemini Context Cache and web-search grounding |
| Groq | Cloud | API key, very fast inference |
| OpenRouter | Cloud | Hundreds of models behind one key |
| Claude (subscription, beta) | Local CLI | Uses your Claude Pro/Max login through the claude CLI, no API key (guide) |
| Custom | Any | Your own endpoint, SSE or NDJSON streaming |
Translation (5)
DeepL, Google, Microsoft, any configured LLM, and OPUS-MT (offline, local models).
How NexQ compares
| NexQ | Otter.ai | Granola | Krisp | |
|---|---|---|---|---|
| Price | Free | Paid tiers | Paid tiers | Paid tiers |
| Open source | Yes | No | No | No |
| Fully local option | Yes | No | No | Partial |
| No bot joins the call | Yes | No | Yes | Yes |
| Choice of STT / LLM | 10 / 9 | Fixed | Fixed | Fixed |
| Local LLM | Yes | No | No | No |
| Your own documents as context | Yes | No | No | No |
- Nothing leaves your machine unless you choose a cloud provider. With a local STT engine and a local LLM, NexQ works offline.
- Your data stays on disk you control: meetings, transcripts, recordings, models and the RAG index live in the app data folder (
%APPDATA%\com.nexq.appon Windows,~/Library/Application Support/com.nexq.appon macOS). - Secrets go to the OS credential store: Windows Credential Manager or the macOS Keychain.
- No telemetry or analytics. Apart from the providers you configure, NexQ only contacts GitHub Releases to check for updates, and every update is verified with an Ed25519 signature.
- Found a security issue? Please follow SECURITY.md.
NexQ is a Tauri 2 app: a Rust backend (tokio) and a React 18 + TypeScript frontend.
Prerequisites
| Windows | macOS (Apple Silicon) | |
|---|---|---|
| Toolchain | Node.js 20+, Rust stable, WebView2 (ships with Windows 11) | Node.js 20+, Rust stable, Xcode Command Line Tools (xcode-select --install) |
| Native build tools | CMake, Ninja, LLVM: winget install Kitware.CMake Ninja-build.Ninja LLVM.LLVM |
CMake: brew install cmake |
git clone https://github.com/naxhq/NexQ.git
cd NexQ
npm install
npx tauri dev # run the full app with hot reload| Task | Command |
|---|---|
| Frontend tests (Vitest) | npm test |
| Rust tests | cargo test --lib (in src-tauri/) |
| Type check | npx tsc --noEmit |
| Production frontend build | npm run build |
| Windows installer | npx tauri build |
| macOS app + dmg | ./build-mac.sh |
Read docs/DEVELOPMENT.md for setup details, debugging and logs, and docs/ARCHITECTURE.md for how the pieces fit together.
Contributions of every size are welcome: bug reports, docs, translations, new providers, and macOS testing in particular. Start with CONTRIBUTING.md and look for good first issue and help wanted. Please follow our Code of Conduct.
- Questions and ideas: GitHub Discussions
- Bugs and feature requests: Issues. For bugs, include your log: Settings → About → Diagnostics → Copy recent log (how)
- Security reports: privately, via SECURITY.md
- More in SUPPORT.md
If NexQ helps you, a star helps others find it.
MIT © Vahid Alizadeh and NexQ contributors.
NexQ stands on the shoulders of
Tauri,
React,
whisper.cpp and whisper-rs,
ONNX Runtime and the ort crate,
sherpa-onnx,
cpal,
rusqlite,
Zustand,
Tailwind CSS and shadcn/ui,
and OPUS-MT.
Thanks to every contributor, including @visitek for the macOS port.