natively-cluely-ai-assistant

Sponsored by Recall AI - API for desktop recording

If you’re looking for a hosted desktop recording API, consider checking out Recall.ai, an API that records Zoom, Google Meet, Microsoft Teams, in-person meetings, and more.

---

The Free-for-Personal-Use, Source-Available Cluely Clone

Natively started as a pixel-perfect recreation of Cluely’s interface — then kept going. If you’ve used Cluely, you already know how to use Natively. Same overlay, same workflow, same shortcuts. Except it’s free for personal, educational, research, and non-commercial use, source-available, runs locally, supports any LLM, and has never breached a single user’s data.

Looking for a free personal-use Cluely alternative? A source-available Cluely clone? You found it.

---

What Users Are Saying

“Natively is significantly faster than Cluely when it comes to response time and screen analysis. The latency is practically non-existent.”
— Premium User

“Just wanted to say thanks! Natively helped me completely crack the first two rounds of my Software Engineering interviews. The responses were incredibly fast and accurate.”
— Private Email Feedback

“Used the free version of Natively for my interviews and just landed a massive summer internship. It took all the stress out of the live coding and behavioral rounds!”
— Private Email Feedback

---

Why Natively?

While other tools act as simple API wrappers, Natively is a complete, native intelligence system designed specifically for high-stakes meetings and interviews.

• Native Audio Capture (<500ms): Built with Rust and Zero-Copy ABI transfers, bypassing generic web-audio limitations for ultra-low latency.
• Local Whisper STT (On-Device): 100% on-device speech-to-text using optimized ONNX models (Moonshine-tiny, Moonshine-base, Whisper-large-v3-turbo, distil-large-v3, Parakeet CTC 0.6B, and Parakeet TDT 0.6B v3 for 25 languages). Uses hardware acceleration (CoreML/Metal GPU on Apple Silicon, DirectML on Windows, quantized int8 on CPU) with zero cloud fees or data exposure.
• Apple Speech STT (On-Device): On macOS 26, Natively can transcribe through Apple’s SpeechAnalyzer and SpeechTranscriber without an API key or sending audio to a speech provider.
• Dual-Channel Intelligence: Distinct pipelines for system audio (what they say) and your microphone (what you dictate) ensuring perfect transcription without room noise.
• The Most Advanced Stealth We Compare: Hidden from screen shares, the Dock and the taskbar, disguised in the process list, with tooltips, dialogs and notifications kept out of the share too. Hardened release after release, and best-effort like any stealth.
• Auto Answer (Beta): Answers appear on their own the moment the other person finishes a question — Natively decides whether an answer is actually wanted and stays silent when it isn’t. Off by default.
• Visual Answers: System designs, flows and charts, drawn live inside the answer, in any mode.
• Phone Mirror 2.0: Follow the meeting from your phone, with the live transcript, answers as they stream in, and photos sent back to the desktop.
• Notepad: Type your own notes while Natively listens, run recipes like “What did I miss” or “List my todos”, and Ask questions that read the whole meeting.
• Your Own Overlay Buttons: Describe a button in plain sentences and Natively turns it into an overlay action beside the built-in ones.
• Connectors and Share Links: Send notes to Gmail, Slack, Notion, Zapier, webhooks and your CRM, share a meeting as a link, and get a follow-up email written in your own style.
• Modes Manager (9 Built-in Personas): Toggle between General, Technical Interview, Looking for Work, Sales, Recruiting, Team Meet, Lecture, Seminar (strict file-grounded Q&A), and Call Center — each with its own system prompt and meeting-note template. Build your own on top.
• Browser Companion (Chrome): Press ⌘/Ctrl+Y to pull the page you’re looking at — a job description, a coding problem, documentation — straight into your next answer.
• Custom Context & Notes: A dedicated free-form notes area to paste instructions, crib sheets, or credentials (up to 8,000 characters), automatically injected into real-time LLM prompts.
• Rolling Context: We don’t just transcribe; we maintain a “memory window” of the conversation for smarter answers.
• Local RAG Memory: We embed your meetings locally using SQLite vector search so you can ask, “What did John say about the API last week?”
• Embeddings and Rerankers, Your Choice: Gemini, OpenAI, Voyage AI, Jina AI or local models decide what your search finds.
• Reference Files: Deeply integrate PDFs, DOCX, and TXT files as real-time context.
• Rich Dashboard: A full UI to manage, search, and export your history—not just a floating window.
• Fully Offline Capable: Don’t trust the cloud? Run Natively 100% offline using local Ollama models and local Whisper STT.

---

4 things you should know before choosing an interview AI

1. Cluely had a data breach in mid-2025 that exposed 83,000 users’ personal info, transcripts, and screenshots. Natively keeps your transcripts, screenshots and keys on your machine and has never had a breach.
2. Final Round AI costs $149/month and its taskbar icon is visible to proctoring software. Natively is free for personal/non-commercial use, source-available, and includes its full stealth feature set at no cost.
3. LockedIn AI charges $55 to $70 a month and locks you into their cloud LLM with no local option. Natively lets you use any model (GPT, Claude, Gemini, Llama) or go fully offline with Ollama.
4. Interview Coder is a paid tool for the coding round only. Natively disguises its process on macOS and Windows, keeps tooltips, dialogs and notifications out of screen shares, covers behavioral and system design rounds too, and is free for personal use.

---

---

Demo

This demo shows a complete live meeting scenario:

• Real-time transcription as the meeting happens
• Rolling context awareness across multiple speakers
• Screenshot analysis of shared slides
• Instant generation of what to say next
• Follow-up questions and concise responses
• All happening live, without recording or post-processing

---

The Stealthiest Free Interview Copilot: Why Natively Leads in 3.0

With 2.9, Natively stopped being “the free Cluely clone” and became the tool to beat. 3.0 adds a notepad and the tools that carry your notes out of the meeting. Five things set it apart from Interview Coder, Cluely, Final Round AI and LockedIn AI.

1. The most advanced stealth of any tool we compare

Other tools hide a window. Natively hides the whole app, at every layer a screen share, a task switcher or a process list can see:

• Invisible in screen shares and recordings. Every Natively window is marked private, so Google Meet, Teams, QuickTime and Zoom (with window filtering) leave it out.
• Nothing it opens leaks. Tooltips, dialogs, dropdowns, file pickers and notifications are separate windows, and an overlay that forgets them shows up in the share. In Undetectable mode Natively keeps every one of them out.
• No Dock icon, no taskbar button, no tray icon while Undetectable mode is on.
• Process Disguise on macOS and Windows. Have the app present itself as Terminal, System Settings or Activity Monitor in the process list, the Dock and the taskbar.
• Type without losing focus (Windows). Stealth typing sends your keys to the overlay while your call keeps focus, and Protect Natively shortcuts stops a shortcut from also typing into the app underneath.
• Quiet on the network. Third-party analytics and crash reporting switch off for as long as Undetectable mode is on.

Interview Coder built its name on process-level hiding. Natively disguises its process on both macOS and Windows, adds everything above, covers the whole interview instead of only the coding round, and is free for personal use.

Stealth is best-effort, not a guarantee. Dedicated proctoring software (Pearson VUE, ProctorU, Respondus Lockdown Browser) is a different category, and a phone camera sees everything. See Is stealth mode actually undetectable?

2. Visual Answers: the system design round, drawn for you

Ask for a system design, a flow or a chart and Natively draws it live inside the answer, while the explanation is still being written.

• Flowcharts, sequence diagrams, ER diagrams, state machines, swimlanes, trees and timelines
• Charts built from the numbers in the conversation, and back to a table when you ask
• “Draw what we discussed” turns the last ten minutes of the meeting into a diagram
• Works in all nine modes, and understands requests in English, Spanish, Russian, Chinese and Japanese

A wall of text explains an architecture. A diagram shows that you understand it.

3. An interface built to feel native

Natively is designed to feel like it belongs on your Mac, and the same design runs on Windows:

• Liquid Glass controls across Settings, the overlay and meeting notes
• Motion with a purpose: cards and popups open with a genie motion, panels fold instead of jumping, and answers stream in word by word
• A first run that teaches itself: a welcome screen and a three-step tour where you try the real shortcuts
• Light and dark themes, a new startup splash, and a model picker that lists models by provider

In our opinion it is the best-looking and smoothest app in the category. Try it for five minutes and judge for yourself.

4. New in 3.0: notes that keep up and leave the meeting with you

Natively 3.0 adds a notepad and the tools that carry your notes out of the meeting:

• Notepad: type your own notes while Natively listens, with a live transcript beside them. Recipes such as “What did I miss”, “Write tldr” and “List my todos” run from a chip or from “/” in Ask, and Ask reads the whole meeting once it ends.
• Your own overlay buttons: describe a button in plain sentences (what it does, what to read, how the answer looks) and it joins the overlay beside Recap and Follow-up questions. Pick the five buttons the overlay shows.
• Connectors: Gmail, Slack, Notion, Zapier, webhooks, HubSpot, Salesforce, Attio, Affinity and Pipedrive, with browser sign-in and auto-send rules, plus a local MCP server for your own tools. Notes go from your computer to the service, never through our servers.
• Share notes as a link: one link to a meeting’s notes and transcript.
• Follow-ups in your style: a compose sheet writes the follow-up email in your tone when you click, set per mode.

5. Free where the others charge

Final Round AI is $149 a month. LockedIn AI is $55 to $70 a month. Cluely starts at $20 a month and sells screen-share invisibility as a separate, higher-priced tier. Interview Coder is paid. Natively is free for personal use with your own keys, and every stealth feature above is included.

---

Full Comparison: Natively vs Cluely vs Final Round AI vs LockedIn AI vs Interview Coder

Feature	Natively	Cluely	Pluely	LockedIn AI	Final Round AI
Price	✅ Free (BYOK)	⚠️ $20/mo	✅ Free	❌ $55–70/mo	❌ $149/mo
Source license	Personal-use source license	❌	✅	❌	❌
Local data / private	✅ Yes	❌ Cloud servers	✅ Yes	❌ Cloud servers	❌ Cloud servers
Any LLM (BYOK)	✅ Yes	❌ Vendor-locked	⚠️ Limited	❌ Vendor-locked	❌ Vendor-locked
Local AI (Ollama)	✅ Yes	❌	❌	❌	❌
Local Whisper (On-Device)	✅ Yes	❌	❌	❌	❌
Real-time <500ms	✅ Yes	⚠️ 5–90s lag	✅ Yes	✅ ~116ms	⚠️ Slowest
Dual audio channels	✅ System + Mic	❌ Single stream	❌	❌	❌
Local RAG memory	✅ SQLite + sqlite-vec	❌	❌	❌	❌
Meeting history	✅ Full dashboard	⚠️ Limited	❌	❌	⚠️ Limited
Screenshot OCR	✅ Yes	⚠️ Limited	❌	✅ Yes	⚠️ Limited
Stealth mode	✅ Undetectable mode, free	⚠️ Paid tier	❌	❌	❌ Visible to proctors
Process Disguise	✅ Terminal, Settings, etc	❌	❌	❌	❌
Resume & context	✅ Pro	❌	❌	✅ Yes	✅ Yes
Custom Personas/Modes	✅ Pro	✅ Yes	❌	❌	⚠️ Limited
Custom Context & Notes	✅ Pro	❌	❌	❌	❌
Multi-Key API Pools	✅ Yes	❌	❌	❌	❌
Profile Intel Router	✅ Yes	❌	❌	❌	❌
Eager Code Expansion	✅ Yes	❌	❌	❌	❌
Live Follow-Up Resolver	✅ Yes	❌	❌	❌	❌
Real-Time Latency Trace	✅ Yes	❌	❌	❌	❌
Phone Mirror 2.0	✅ Yes	❌	❌	❌	❌
Auto-Calendar Sync	✅ Yes	❌	❌	❌	❌
Smart Task Sync	✅ Yes	❌	❌	❌	❌
Speaker Diarization	✅ Yes	❌	❌	❌	❌
Room Speakers (shared mic)	✅ Yes	❌	❌	❌	❌
Notepad + Recipes	✅ Yes	⚠️ AI notes	❌	❌	❌
Your Own Overlay Buttons	✅ Plain sentences	❌	❌	❌	❌
Connectors (Gmail, Slack, Notion, CRMs)	✅ 11 + webhooks	⚠️ Slack, Notion	❌	❌	❌
Local MCP Server	✅ Yes	❌	❌	❌	❌
Hands-Free Auto Answer	✅ Beta	❌	❌	❌	❌
Browser Page Capture	✅ Chrome extension	❌	❌	❌	❌
JD Role Insight	✅ Pro	❌	❌	⚠️ Limited	⚠️ Limited
9 Built-in Modes	✅ Yes	⚠️ Limited	❌	❌	⚠️ Limited
Codex CLI Integration	✅ Yes	❌	❌	❌	❌
Offline SLM Mode	✅ Yes	❌	❌	❌	❌
Stateful Intelligence OS	✅ Yes	❌	❌	❌	❌
Hindsight LTM Vector DB	✅ Yes	❌	❌	❌	❌
Regional STT Relay	✅ Yes	❌	❌	❌	❌
Data breach history	✅ None	❌ 83k users exposed	✅ None	✅ None	✅ None

Legend: ✅ Full support · ⚠️ Partial or limited · ❌ Not available

---

Why Natively wins

vs Cluely — breached 83,000 users

The UI is intentionally familiar — if you’ve used Cluely, there’s zero learning curve.

Cluely’s mid-2025 data breach exposed personal information, full interview transcripts, and screenshots of 83,000 users. Every word spoken during an interview was stored on their servers — and then leaked. They charge $20/month for this privilege.

By default, Natively stores your transcripts, API keys and screenshots on your local machine, and they never leave it when you use your own keys. The app reports feature-usage statistics that never contain what you say, see or type, and you can switch them off. The codebase is source-available under the Natively Personal Use Source License v1.0 and auditable. Zero breaches. That is the only acceptable standard for a tool that listens to your interviews.

Unlike Cluely’s rigid interface, Natively also gives you complete control over the AI: Custom Persona Modes (Tech, Sales, Recruiting) to strictly format behavior, and Reference Files capabilities to upload PDFs so the AI knows exactly the context of the job or meeting before it starts.

Cluely also sells its screen-share invisibility as a separate, higher-priced tier. In Natively, Undetectable mode is part of the free app.

vs LockedIn AI — $70/month for cloud lock-in

LockedIn AI is the most expensive tool in the category at $55–70/month. It locks you into a single cloud LLM with no option for local inference. Every transcript and response passes through their servers.

Natively supports every major model (Gemini, GPT, Claude, Groq) via bring-your-own-key, and offers 100% offline mode through Ollama. You pay only for the API tokens you actually use — or pay nothing at all by running Llama 3 locally. No subscription, no vendor lock-in.

vs Final Round AI — $149/month and visible to proctors

Final Round AI is the most expensive option at $149/month, optimized for pre-interview prep and mock interviews but with the slowest live latency in the category. Critically, its taskbar icon is visible to proctoring software, making it detectable during monitored interviews.

Natively delivers <500ms end-to-end latency using Rust-based native audio capture with Zero-Copy ABI Transfers. Its Undetectable mode hides the app from screen shares, the Dock and the taskbar, disguises its processes, and has been hardened release after release.

vs Pluely — lightweight but limited

Pluely is a solid lightweight alternative (~10MB, Tauri-based) and it has Linux support, which Natively does not yet offer. Credit where it is due.

But Pluely is a basic overlay. It has no local RAG, no meeting history, no dual audio channels, and no dashboard. Natively is a complete intelligence system: it remembers your past meetings via local vector search, separates system audio from your microphone, and gives you a full management dashboard with export to Markdown, JSON, and Text.

vs Interview Coder: Matching Stealth, More Features, Completely Free

Interview Coder is a paid tool focused on the coding round, and process-level hiding is its selling point. Natively disguises its process on macOS and Windows too, then adds what a whole interview needs, for free:

	Natively	Interview Coder
Price	✅ Free (BYOK)	❌ Paid
Source license	Personal-use source license	❌
Works on LeetCode / HackerRank	✅	✅
Screenshot + OCR analysis	✅	✅
Real-time overlay	✅	✅
Process-level disguise	✅ macOS + Windows	✅
Local AI / offline mode	✅ Ollama	❌
Behavioral interview support	✅	❌
System design support	✅	❌
Meeting history & RAG	✅	❌
Meeting notepad & connectors	✅	❌
Any LLM (BYOK)	✅	❌ Locked
Data stored locally	✅	❌ Cloud

Natively covers the full interview loop, not just the coding round: behavioral answers drawn from your résumé, system designs drawn live with Visual Answers, and notes when the call ends.

vs Parakeet AI — Memory and History vs Stateless Overlay

Parakeet AI offers basic live meeting assistance but has no persistent memory, no meeting history, and no local vector search. Natively remembers your past meetings via local RAG, lets you ask questions across all your history, and gives you a full dashboard to manage, export, and search everything. Furthermore, Natively includes Custom Persona Modes allowing the AI to structure notes and behave optimally for specific flavors of conversations, instead of relying on Parakeet’s one-size-fits-all model.

---

Where we’re not there yet

• No Linux support — we are actively looking for maintainers to help bring Natively to Linux
• API key setup overhead — you need to bring your own API keys (or install Ollama), which adds initial setup friction compared to all-in-one cloud tools
• No built-in mock interview mode — Final Round AI has dedicated mock interview practice; Natively focuses on live, real-time assistance

---

Free AI Coding Interview Assistant — Undetectable on LeetCode, HackerRank & CoderPad

Natively works as a free, undetectable AI coding interview assistant for standard online assessments. It captures your screen, analyzes the problem, and gives you real-time hints, solutions, and explanations — all through an invisible overlay that doesn’t interfere with your coding environment.

Works undetected on:

• LeetCode (including LeetCode contests)
• HackerRank
• CoderPad
• Codility
• HackerEarth
• Karat
• Any browser-based coding environment

How it works:

1. Screenshot the problem with a single shortcut
2. Natively OCRs the question and sends it to your chosen AI (GPT, Claude, Gemini, or local Ollama)
3. Response appears in the invisible overlay — never on screen share

⚠️ Important: Natively is not designed to bypass dedicated proctoring software like Pearson VUE, ProctorU, or Respondus Lockdown Browser. These run at the OS level and are a different category entirely. For standard online coding assessments without dedicated proctoring software, Natively’s stealth mode is built to stay out of sight, on a best-effort basis.

---

---

Natively API (Hosted Tier)

Stop managing four separate services. One key. Zero configuration.

Are you managing separate accounts for your AI reasoning, live transcription, fast inference, and web search? Juggling multiple API keys, rate limits, and invoices across completely different categories of tools is unnecessary overhead. Natively API replaces all of those categories with one flat subscription.

Under the hood, Natively API connects you to the absolute best models for the optimal user experience:

• Backend AI Models: Claude, OpenAI, Gemini, and Groq.
• Premium STT Models: Google Chirp 2/3, ElevenLabs Scribe v2, and Deepgram Nova-3.

4 Categories → 1 Key

Your current unbundled stack:

• AI Intelligence (GPT/Claude/Gemini): per-token billing and usage anxiety
• Lightning-Fast Inference (Groq/Llama): strict rate limits to monitor
• Real-Time Transcription (Deepgram/Google STT): separate key + quota
• Web Search & Research (Tavily/Perplexity): yet another subscription

Replaced by Natively API:

• AI chat, transcription & web search — all included
• One flat subscription. Zero surprise bills. Starts at $8/mo.
• Single key. Zero rotation. Zero configuration.

API Plan Comparison

Feature	Standard ($8/mo)	Pro ($15/mo)	Max ($25/mo)	Ultra ($35/mo)
All-in-One Cloud AI Access	✅ Yes	✅ Yes	✅ Yes	✅ Yes
Real-Time Transcription	✅ Yes	✅ Yes	✅ Yes	✅ Yes
Included Natively Pro Desktop App	❌ No	✅ Yes	✅ Yes	✅ Yes
Premium Support	❌ No	✅ Yes	✅ Yes	✅ Yes
Higher Monthly Quotas	❌ No	✅ Yes	✅ Yes	✅ Yes

Don’t start the long way. Skip the 20-minute manual setup. One Natively subscription skips all of it — AI, transcription, and web search are ready immediately.

---

Natively Pro

While Natively is free for personal, educational, research, and non-commercial use, we also offer a Pro Edition (available as Lifetime or Yearly subscriptions) designed specifically for power users and job seekers. Purchasing a Pro license gives you an edge in the job market, all while directly supporting the continued development of Natively.

Free vs Pro Feature Comparison

Feature	Natively Free	Natively Pro
Bring Your Own Key (BYOK) Models	✅	✅
Local AI Support (Ollama)	✅	✅
Local Whisper STT (On-Device)	✅	✅
Real-Time Speech-to-Text (<500ms)	✅	✅
Multi-Key API Pools & Key Rotation	✅	✅
Profile Intelligence Router (v3)	✅	✅
Auto Answer (Beta)	✅	✅
Browser Companion & Page Capture (Chrome)	✅	✅
9 Built-in Modes (incl. Seminar, Call Center)	✅	✅
Skills — Upload & Manage Custom Personas	✅	✅
Meeting Notes with Follow-Up Email Draft	✅	✅
Phone Mirror 2.0	✅	✅
Notepad with Recipes & Live Transcript	✅	✅
Your Own Overlay Buttons	✅	✅
Connectors (Gmail, Slack, Notion, CRMs, MCP)	❌	✅
Share Notes as a Link	✅	✅
Follow-Up Emails in Your Style	✅	✅
Room Speakers (People Sharing One Mic)	✅	✅
Calendar Sync (Google Calendar)	✅	✅
Visual Answers (Live Diagrams & Charts)	✅	✅
Eager Code UI Expansion	✅	✅
Live Follow-Up Resolver	✅	✅
Real-Time Latency Tracing	✅	✅
Two New Meeting UI Styles (Liquid Glass/Modern)	✅	✅
Live Contextual Assistant	✅	✅
Screenshot & Slide OCR Analysis	✅	✅
Undetectable & Stealth Modes	✅	✅
Meeting Dashboard & Offline RAG History	✅	✅
Stateful “Intelligence OS”	✅	✅
Hindsight Long-Term Memory (LTM)	❌	✅
Job Description (JD) & Resume Context Awareness	❌	✅
Automated Company Research & Dossiers	❌	✅
Role Insight — JD Requirement Coverage Scoring	❌	✅
Live Salary & Offer Negotiation Copilot	❌	✅
Custom Persona Modes (Sales, Tech, etc.)	❌	✅
Custom Context & Notes	❌	✅
Reference Files (PDF/DOCX/TXT upload)	❌	✅
Speaker Diarization	❌	✅
Priority Feature Access & Support	❌	✅

What’s New in v3.1.0 (Latest Release)

Pick up where you left off. Resume a saved meeting, a chat that carries on after Resume, a Compare that clears the losing version, and a 9-minute setup film.

New

Feature	What it does
Resume a meeting	Record more into a meeting that has already ended, and it stays one meeting. Resume is on the meeting page for overlay meetings and in the notepad for notepad meetings, and at Stop the notes are rewritten over the whole meeting.
Chat carries on after Resume	The overlay reopens with the meeting’s earlier answers, so the conversation picks up where it stopped.
Same speaker after Resume	Someone who talks again after a Resume can come back as a new Speaker N. The meeting page asks “Same as Lars?” and one click gives them the saved name.
Open a meeting in the notepad	A notepad button on the meeting page opens the notepad on that saved meeting, with its notes.
Setup & demo film	A 9-minute film from setup to a live meeting, from the Welcome screen, Settings → Setup & Help or the card on Home.

Improved

• Compare. Picking a winner deletes the losing version and its tab. An older winner becomes the meeting’s notes, and “About the same” keeps both.
• Free trial. The trial ends when its AI allowance runs out as well as at 30 minutes, warns about two answers before, and a refused answer says the trial ended instead of “check your API keys”.
• Backup speech. When your provider stops your key for quota, the banner says so and names the provider transcribing in the meantime.
• Local speech models say why they stopped: not enough memory (with the measured figures), not downloaded, or stopped.
• Pro lifetime is $60, with $79 struck through during a named sale.
• Settings. Ambient AI Chat is back in General, and Do not save meetings moves under Advanced. Meeting notes fade out under the pinned header.

Fixed

• Speech language. Auto follows the speaker for the whole meeting on Natively speech, instead of locking onto the first language it heard.
• Garbled transcripts. English no longer comes back as nonsense in another alphabet on microphones whose audio could not be converted to 16 kHz.
• Windows notepad. The notes window takes clicks again after a second notepad meeting.
• Overlay. A banner whose text changed twice quickly no longer goes blank, and light-mode code blocks lose the white bar behind each line.
• Free trial. The end-of-trial card shows what the trial used instead of zeros.
• Local models. A memory reading the OS fails to take no longer blocks every local model.
• Debug log. Opening Natively while it is already running no longer wipes the running app’s log.

Platform

• The native audio module builds on macOS 27 with Xcode 27, and the shipped module is stripped as before on macOS and Windows.
• Electron 43.7.7 and Node 22.13 or later, unchanged from 3.0.0.

📄 Read the full v3.1.0 release notes → · v3.0.0 release notes

Table of Contents

• The free personal-use Cluely clone
• What Users Are Saying
• Why Natively?
• 4 things to know
• Why Natively leads in 3.0
• Demo
• Full comparison
• Why Natively wins
• AI Coding Assistant
• Natively Pro
• What’s New in v3.1.0
• Privacy & Security
• Installation
• AI Providers
• Key Features
• Meeting Intelligence Dashboard
• Roadmap
• Use Cases
• Technical Details
• Known Limitations
• Responsible Use
• Contributing
• License
• FAQ
• Alternatives Natively replaces
• Star History

---

What Is Natively?

Natively is a desktop AI assistant for live situations:

• Meetings
• Interviews
• Presentations
• Classes
• Professional conversations

It provides:

• Live answers
• Rolling conversational context
• Screenshot and document understanding
• Real-time speech-to-text
• Instant suggestions for what to say next

All while remaining invisible, fast, and privacy-first.

---

Privacy & Security (Core Design Principle)

• Source-available under the Natively Personal Use Source License v1.0
• Bring Your Own Keys (BYOK)
• Local AI option (Ollama)
• All data stored locally
• Feature-usage statistics that never contain what you say, see or type, with a switch to turn them off (see PRIVACY.md)
• No hidden uploads

You explicitly control:

• What runs locally
• What uses cloud AI
• Which providers are enabled

---

Installation (Developers & Contributors)

Prerequisites

• Node.js (v22.13+ required)
• Git
• Rust (required for native audio capture)
• Xcode 26+ with the macOS 26 SDK (required only to build the Apple Speech helper or package Natively for macOS)

AI Credentials & Speech Providers

Natively is 100% free to use with your own keys.
Connect any speech provider and any LLM. No subscriptions, no markups, no hidden fees. All keys are stored locally.

Speech-to-Text Providers

• Apple Speech (No API Key) - On-device transcription on macOS 26+
• Local models (No API Key) - Whisper, Moonshine and Parakeet on your device, including Parakeet TDT for 25 languages and Parakeet v2 for English, which runs on the Neural Engine on Apple Silicon Macs with macOS 14+
• Soniox (API Key) - Ultra-fast, highly accurate streaming STT (recommended)
• NVIDIA Riva / NIM (API Key) - Nine streaming models, including multilingual Nemotron and Parakeet
• Qwen (API Key) - Realtime streaming speech recognition
• Google Cloud Speech-to-Text (Service Account)
• Groq (API Key)
• OpenAI Whisper (API Key)
• Deepgram (API Key)
• ElevenLabs (API Key)
• Azure Speech Services (API Key + Region)
• IBM Watson (API Key + Region)

Apple Speech requires a Mac running macOS 26 or later. Select Apple Speech in Audio settings and choose a recognition language. On first use, macOS may download the corresponding speech asset from Apple; the initial startup can therefore take longer and requires a network connection. Once the asset is installed, transcription runs on the device. Language availability is determined by macOS.

AI Engine Support (Bring Your Own Key)

Connect Natively to any leading model or local inference engine.

Provider	Best For
Gemini 3.1 / 3.8 Series	Recommended (and the default): 2M-token context & low cost.
OpenAI (GPT-5.4 & o3)	High reasoning capabilities.
Anthropic (Claude 4.6)	Coding & complex nuanced tasks.
Groq (Qwen 3.8/Scout 4)	Insane speed (near-instant answers) & screenshot analysis.
NVIDIA NIM	NVIDIA-hosted open models, plus Riva streaming speech.
LiteLLM Gateway	Route through your own proxy with a built-in model picker.
Google Antigravity	Sign in with your Google account instead of pasting a key.
OpenRouter / Sidrune	Hosted gateways to many models.
9Router / AgentRouter	More gateways, alongside LiteLLM.
Ollama / LocalAI	100% Offline & Private (No API keys needed).
OpenAI-Compatible	Connect to any custom endpoint (vLLM, LM Studio, etc.)

Note: You only need ONE speech provider to get started. Use Apple Speech for on-device transcription on macOS 26+, or Google STT, Groq, or Deepgram for fast cloud transcription.

---

To Use Google Speech-to-Text (Optional)

Your credentials:

• Never leave your machine
• Are not logged, proxied, or stored remotely
• Are used only locally by the app

What You Need:

• Google Cloud account
• Billing enabled
• Speech-to-Text API enabled
• Service Account JSON key

Setup Summary:

1. Create or select a Google Cloud project
2. Enable Speech-to-Text API
3. Create a Service Account
4. Assign role: roles/speech.client
5. Generate and download a JSON key
6. Point Natively to the JSON file in settings

---

Development Setup

Clone the Repository

git clone https://github.com/Natively-AI-assistant/natively-cluely-ai-assistant.git
cd natively-cluely-ai-assistant

Install Dependencies

npm install

Build Native Audio Module (Rust)

npm run build:native

Build Apple Speech Helper (macOS Only)

Regular development and non-macOS builds do not require the helper. To use Apple Speech while developing on macOS, build it explicitly for the current target architecture:

npm run build:apple-speech -- --arch arm64

Use x64 for Intel or universal to combine both architectures. This command requires Xcode 26+ with the macOS 26 SDK. Production macOS packaging builds the matching helper automatically.

Environment Variables

Create a .env file:

# Cloud AI
GEMINI_API_KEY=your_key
GROQ_API_KEY=your_key
OPENAI_API_KEY=your_key
CLAUDE_API_KEY=your_key
GOOGLE_APPLICATION_CREDENTIALS=/absolute/path/to/service-account.json

# Speech Providers (Optional - only one needed)
DEEPGRAM_API_KEY=your_key
ELEVENLABS_API_KEY=your_key
AZURE_SPEECH_KEY=your_key
AZURE_SPEECH_REGION=eastus
IBM_WATSON_API_KEY=your_key
IBM_WATSON_REGION=us-south

# Local AI (Ollama)
USE_OLLAMA=true
OLLAMA_MODEL=llama3.2
OLLAMA_URL=http://localhost:11434

# Default Model Configuration
DEFAULT_MODEL=gemini-3.1-flash-lite-preview

Run (Development)

npm start

Build (Production)

npm run dist

This runs: Vite build → TypeScript compile → native module build → electron-builder

---

AI Providers

• Custom (BYO Endpoint): Paste any cURL command to use OpenRouter, DeepSeek, or private endpoints.
• Ollama (Local): Zero-setup detection of local models (Llama 3, Mistral, Gemma).
• Dynamic Model Selection: Preferred models (OpenAI, Anthropic, Google) now automatically appear across the app.
• NVIDIA NIM: NVIDIA-hosted models for both text and streaming speech-to-text.
• LiteLLM: Point Natively at your own gateway, with a model picker and configurable default.
• Google Gemini: First-class support for the Gemini 3.1 and 3.8 series, the default models.
• OpenAI: GPT-5.4 and o3 series support with optimized system prompts.
• Anthropic: Claude 4.6 series support with corrected max_tokens.
• Groq: Ultra-fast text inference with Qwen 3.8, and screenshot analysis using Llama 4 Scout.
• Google Antigravity: Sign in with your Google account to use Gemini models without an API key.
• Gateways: OpenRouter, Sidrune (formerly Fluxion), 9Router and AgentRouter are supported as providers.
• Codex: Uses your codex login session, with GPT-6 Astra and GPT-5.6 Sol as built-in models.
• Background Model: A second, faster model for Auto Answer, quick decisions and Fast Response Mode.
• Provider fallback: When your model fails or stalls, Natively retries it, moves to your next provider, and tells you which one answered.
• Reads images: An Auto, On or Off setting per model decides which ones receive screenshots.

---

Key Features

Invisible Desktop Assistant

• Always-on-top translucent overlay
• Instantly hide/show with shortcuts
• Restoring with Toggle Visibility (Cmd+B / Ctrl+B by default) brings Natively to the display under the mouse pointer. It centers when changing displays and keeps its previous position on the same display, adjusting the window to fit the usable area. Screenshot and automatic restores keep their existing placement.
• Works across all applications

Real-time Interview Copilot & Coding Help

• Real-time speech-to-text (<500ms latency)
• Fast Response Mode: Quicker text responses from your Background Model, or from the fastest available provider.
• Multilingual Support: Choose from various response languages, and set speech recognition matching specific accents and dialects.
• Anti-Chatbot / Human Persona System: Refined system prompts and negative constraints ensure responses are concise, conversational, and indistinguishable from a real candidate (no robotic preambles or lectures).
• Context-aware Memory (RAG) for Past Meetings
• Instant answers as questions are asked
• Interim/Final Bridging: Manual transcript finalization and interim bridging during recordings for higher accuracy.
• Smart Recap & Summaries: Instant meeting minutes and executive summaries.
• TinyPrompts™ Engine: Specialized prompt architecture for local SLMs (4B-8B params), ensuring instruction following and reasoning parity with cloud models on local hardware.
• Dynamic Note Templates: AI automatically generates structured meeting notes based on your active persona mode (e.g., Tech Interview follow-ups vs Sales action items).

Auto Answer (Beta) — Hands-Free Responses

Natively can answer without being asked. When the other person finishes a question, an LLM judge decides whether it is genuinely a question aimed at you — and either answers it or stays silent.

• No prompt, no button. Nothing to dismiss and nothing to click; the answer is simply there.
• Knows when to stay quiet. Rhetorical asides, questions aimed at someone else, and restatements of something already answered are ignored.
• Speaker-aware. It reads speaker labels and splits the conversation by voice, so your own thinking-out-loud doesn’t trigger an answer.
• Off by default. Turn it on in Settings → General.

Visual Answers: Diagrams and Charts in the Answer

Ask for a picture and the answer draws one, live.

• System designs, flows and timelines, drawn while the answer is written
• Charts from numbers, and “show the chart as a table” turns one back into a table
• Drawings for every mode: flowcharts, sequence diagrams, ER diagrams, state machines, swimlanes and trees
• “Draw what we discussed” reads the last ten minutes of the meeting transcript
• Five languages: requests are understood in English, Spanish, Russian, Chinese and Japanese

Browser Companion — Capture What You’re Reading

A Chrome extension that hands the page you’re on to Natively.

• ⌘/Ctrl+Y pulls the current page into your next answer — job descriptions, coding problems, documentation, dashboards
• One-click permissions for every site, instead of approving them one at a time
• Automatic context recognises coding pages, job descriptions and developer docs and captures them without being asked
• Secrets are stripped from captured page content before it ever reaches a model
• Falls back to a screenshot, with a visible notice, when a page can’t be read

Instant Screen & Slide Analysis (OCR) — AI Coding Interview Assistant

• Works on LeetCode, HackerRank, CoderPad, Codility, HackerEarth and any browser-based coding environment
• Capture a coding problem with one shortcut — get a full solution, explanation, and complexity analysis instantly
• Eager Code Expansion: Overlay dynamically resizes to accommodate incoming code blocks before React mounts the markdown code rows, preventing visual layout jumps.
• Hardware-Accelerated Transitions: Polished, custom cubic-bezier tweens handle UI growth smoothly, preserving candidate stealth and presentation quality.
• Invisible overlay never appears on screen share or recordings
• Multiple screenshot support for multi-part problems
• Reads images: each model has an Auto, On or Off setting for screenshots, and your selected model reads your screenshot first when it can
• Smart fallback to another vision model if the first one fails

Premium Profile Intelligence

• Profile Intelligence Router (v3): Categorizes each question into a domain (Coding, System Design, Behavioral, Negotiation) and picks its source deliberately — your résumé, an uploaded document, the live conversation, or general knowledge — rather than blending them.
• Role Insight: Paste a job description and see every requirement extracted, assessed against your profile, and scored for coverage before the interview starts.
• Answer-Type Constraints & Follow-Up Resolver: Contextually tracks conversations to answer subsequent queries, and enforces precise layout constraints (such as short, conversational, bulleted, or code-only responses).
• Custom Persona Modes: Seamlessly switch between built-in personas (Technical Interview, Sales, Recruiting) or create your own custom modes tailored to any conversation.
• Reference Files & Custom Context: Upload PDFs, DOCX files, or type custom instructions to give the AI real-time context on your specific situation.
• Job Description & Resume Context: Natively understands your background and the role you’re applying for to provide highly tailored, context-aware answers.
• Company Research: Get instant intelligence and dossiers on the company you are interviewing with.
• Negotiation Assistance: Real-time guidance and strategy during offer and salary negotiations.
• Evidence Validator & Live Deadlines: Real-time validation of factual claims and interactive deadline alert tracker during live assessments.
• PI Latency Tracer: Built-in granular latency profiling mapping exact time spent during the routing and LLM inference loop.

Skills — Custom AI Personas

Create local SKILL.md files to give the AI specialized instructions for any task. Skills are invoked directly from the overlay chat:

• Type / or $ to open a live skill picker — filtered autocomplete with arrow-key navigation, just like Claude Code’s slash commands
• Or type /skill-name directly to activate a skill inline
• Built-in: Humanize AI Text — strips AI writing patterns and makes output sound human
• Upload from the app: drop in a SKILL.md file or a whole skill folder and preview what it does before it installs — no manual file copying
• Delete from the app: remove any skill you installed, straight from the same panel
• Or add one by hand: a SKILL.md with YAML frontmatter name: and description: in ~/Library/Application Support/natively/skills/<folder>/

Contextual Actions

• What should I answer?
• Shorten response
• Recap conversation
• Suggest follow-up questions
• Manual or voice-triggered prompts
• Your own buttons: describe a button in plain sentences (what it does, what to read, how the answer looks) and Natively turns it into an overlay action. Up to four of your own, and you pick the five buttons the overlay shows, in Settings → Overlay.
• Images: paste or drop an image into the overlay, and screenshots and page captures stay in context across hotkeys, quick actions and typed chat.

Seamless Integrations & Sync

• Phone Mirror 2.0: Follow the meeting from your phone, with the live transcript, answers as they stream in, and photos sent back to the desktop. Set it up in Settings → Companion.
• Calendar Sync: Connect Google Calendar and each meeting is linked to its event, notes name the people who spoke, and a follow-up email is drafted when you ask.
• Connectors (Pro): Send notes to Gmail, Slack, Notion, Zapier, webhooks, HubSpot, Salesforce, Attio, Affinity and Pipedrive, with browser sign-in and auto-send rules. A local MCP server and API let your own tools read your meetings. Notes go from your computer to the service, never through our servers. In Settings → Connectors. Connectors are part of Natively Pro (the free trial includes them). Their code lives in this repo, not the premium module, so you can read it, report issues and send pull requests for any connector.
• Share Notes as a Link: Send anyone a link to a meeting’s notes and transcript, from the meeting’s Share menu.
• Knows Your Meeting App: Reads Zoom and Teams for the meeting ID, the people on the call and whether you are muted, and follows the microphone the meeting app uses.
• Live Model Catalogue: Each provider’s model list is fetched live, so new models show up without an app update.
• Smart Task Export: Send extracted action items directly to Jira, Linear, or Asana.
• Speaker Diarization: Real-time speaker identification tags individual speakers by name automatically.
• Room Speakers: People sharing one microphone are told apart in the transcript and notes, on Natively, Deepgram and Soniox speech.
• Codex CLI: Execute terminal tasks, manage workspace files, and run sandboxed code via native Codex integration.

Dual-Channel Audio Intelligence

Natively understands that listening to a meeting and talking to an AI are different tasks. We treat them separately:

• System Audio (The Meeting): Captures high-fidelity audio directly from your OS (fully supported on both macOS and Windows). It “hears” what your colleagues are saying without interference from your room noise.
• Sample Rate Auto-Detection: Dynamically detects and syncs true hardware sample rates (e.g., automatically handling 48kHz audio interfaces or external microphones without distortion or downsampling artifacts).
• Two-Stage Silence Processing: Combines adaptive RMS thresholds with WebRTC Machine Learning VAD to reject typing and fan noise.
• Microphone Input (Your Voice): A dedicated channel for your voice commands and dictation. Toggle it instantly to ask Natively a private question without muting your meeting software.
• Apple Speech (macOS 26+): Processes both channels through Apple’s on-device speech framework, with automatic sample-rate conversion and no speech API key.

Spotlight Search & Customization

• Global activation shortcut (Cmd+K / Ctrl+K)
• Custom Key Bindings: Customize global shortcuts for easier control.
• Instant answer overlay
• Upcoming meeting readiness

Local RAG & Long-Term Memory

• Full Offline RAG: With a local embedding model, all vector embeddings and retrieval happen on your device (SQLite + sqlite-vec).
• Embeddings and Rerankers: Pick Gemini, OpenAI, Voyage AI or Jina AI, or stay on local models. In Settings → Retrieval.
• Semantic Search: innovative “Smart Scope” detects if you are asking about the current meeting or a past one.
• Sliding-Window RAG: 50-token semantic overlap to prevent context loss across chunk boundaries.
• Epoch Summarization: Smarter transcript memory management instead of hard truncation — no more losing early meeting context.
• Global Knowledge: Ask questions across all your past meetings (“What did we decide about the API last month?”).
• Automatic Indexing: Meetings are automatically chunked, embedded, and indexed in the background.

Advanced Privacy & Stealth

• Undetectable Mode: Instantly hide from dock/taskbar with visually locked selector to prevent state mismatches.
• Cross-Window State Sync: Real-time state synchronization across Settings, Launcher, and Overlay windows.
• Process Disguise (Masquerading): Instantly change the app to look like Terminal, System Settings, Activity Monitor, or other harmless utilities to completely evade detection during screen sharing.
• Security Hardening: API keys are scrubbed from memory on app quit and credentials manager overwrites key data before disposal.
• API Rate Limiting: Token-bucket algorithm (burst/refill) to prevent 429 errors on free-tier providers.

README 内容过长，已截取前 64 KB。完整内容请前往 GitHub 仓库查看。