Key Takeaways
- Cluely is a desktop overlay app for macOS and Windows that reads your screen via OCR and listens to your microphone during live video calls, then generates real-time AI-suggested responses visible only to you. It works across Zoom, Google Meet, Microsoft Teams, Webex, and Slack huddles without joining as a visible bot participant.
- Cluely was founded by Chungin “Roy” Lee and Neel Shanmugam after both were suspended from Columbia University in March 2025 for building Interview Coder, a Chrome extension that fed AI-generated coding answers during technical interviews. Rather than walking back the premise, Lee relaunched as Cluely with explicit “cheat on everything” marketing: generating 70,000 signups in the first week after launch.
- The company raised $5.3 million in seed funding and followed that with $15 million from Andreessen Horowitz, reaching a reported valuation of approximately $120 million within months of launch. The funding came despite, or because of, the controversy surrounding the product’s framing.
- Cluely offers a free Starter tier (5 AI responses per day, 100-character output limit), a Pro plan at approximately $20 per month (unlimited responses, latest LLMs), and a Pro plus Undetectability tier at $75 per month that adds a GPU-level stealth overlay invisible to Zoom, Meet, Teams, and Webex screen sharing.
- The stealth overlay on the higher tier uses GPU-level rendering that sits below the layer that standard screen-capture software reads, making the overlay invisible to other participants viewing your screen. However, OS process-level detection tools like Zero Assist can still identify Cluely’s presence in under 500 milliseconds, and dedicated detection products (Validia’s Truely, Proctaroo) are in active development targeting tools of this type.
- In mid-2025, a security breach exposed data on over 83,000 Cluely users, including payment-associated account information, because the company had left admin credentials in a public GitHub repository. Lee publicly acknowledged the breach. The incident is the primary security concern for anyone evaluating the product.
- The practical question for most users is not whether Cluely works technically, its transcription achieves approximately 95% accuracy at around 300ms latency, but whether the risk profile (detection, data handling, terms of service violations with employers and platforms, ethical dimensions) is acceptable for the specific use case.
- Cluely’s most ethically defensible use cases are legitimate ones: real-time meeting notes, live sales call support, live customer service assistance, and language support for non-native speakers in professional meetings. The “cheat on everything” framing dominated early coverage, but the product functions as a general real-time AI meeting assistant for any use case.
No AI product launch in 2025 generated more Reddit discussion, podcast debates, or genuine philosophical argument than Cluely. A tool that literally lets you receive AI answers during live interviews while the person interviewing you has no idea you are getting help cuts to something fundamental about authenticity, fairness, and what skills are actually being tested in modern professional evaluation.
This review covers what Cluely actually is, how it works technically, what it costs, what the controversy was really about, whether the stealth features hold up to scrutiny, and who has a legitimate case for using it. The ethical dimension is real and worth engaging with seriously, delivering not just flagging and ignoring.
What Is Cluely?
Cluely is a real-time AI overlay application that runs on your desktop during video calls. It monitors your screen via OCR and listens to the conversation via your microphone, feeds that context to an AI model, and displays suggested responses in a window visible only to you, not to other participants in the call. It works with Zoom, Google Meet, Microsoft Teams, Webex, and Slack huddles. It never appears as a participant in the call and generates no visible presence that other participants can see through normal video conferencing interfaces.
The product is built specifically around the gap between what a person knows and what they are expected to know in a high-stakes real-time conversation. Job interviews, sales calls, customer service interactions, performance reviews, and professional meetings are all contexts where having access to AI assistance in the moment would change the outcome. Cluely is designed to provide that assistance invisibly.
Responses are hotkey-triggered rather than automatic: you ask Cluely for help when you want it rather than receiving a constant stream of suggestions. The AI has context from both what is appearing on your screen and what it has heard in the conversation so far, making the suggestions contextually relevant rather than generic.
The Origin Story: Columbia, Interview Coder, and “Cheat on Everything”
Cluely’s founding story is inseparable from the product itself. Chungin “Roy” Lee and Neel Shanmugam built Interview Coder while students at Columbia University: a Chrome extension that provided AI-generated coding solutions for LeetCode problems during live technical interviews. Columbia suspended both students in March 2025 for the project.
The standard response to that kind of institutional censure is to apologize, back down, and find a less controversial direction. Lee did the opposite. He launched Cluely on April 20, 2025 with a marketing video that showed him on a date using the product through smart glasses to fake knowledge about art and lie about his age. The video went viral. The explicit “cheat on everything” framing was not accidental positioning; it was the actual pitch.
The launch generated 70,000 signups in the first week. Within five months, Andreessen Horowitz led a $15 million funding round. The company reached a reported valuation of approximately $120 million. The controversy did not slow the growth; for a substantial number of professionals facing high-stakes interviews and calls, the product was solving a real problem they were willing to pay for.
The marketing framing has softened since launch. Cluely now presents itself as an AI meeting assistant with legitimate professional use cases, which is also an accurate description of what it does. The early positioning was built for virality; the actual product is a real-time AI tool that works for any conversation where having AI assistance matters.
How Cluely Works
Cluely runs as a desktop overlay on macOS and Windows, with an iOS app added in 2026. The core technical loop is straightforward: screen OCR reads visual content appearing on your display, microphone input transcribes audio from the call in 12+ languages at approximately 95% claimed accuracy with roughly 300 milliseconds of latency, and the combined context is sent to an LLM that generates a suggested response you see in the overlay window.
The overlay is a transparent window positioned over your screen. On the standard Pro plan, this window appears in your own view but is visible to anyone viewing your screen through screen sharing: the same way any other window would be. On the Pro plus Undetectability tier, Cluely renders through GPU-level hooks that sit below the layer that standard screen-capture software reads. This makes the overlay invisible when you share your screen in Zoom, Meet, Teams, or Webex: other participants see your screen without the Cluely window.
You control when Cluely responds using hotkeys. It does not automatically push suggestions throughout a call; you trigger it when you want help with a specific question or point in the conversation. The AI has running context from the entire session up to that point, so suggestions are grounded in what has actually been said rather than responding to isolated prompts.
Cluely Features
Real-Time AI Responses
The core feature is hotkey-triggered AI suggestions during live conversations. The suggestions are generated from the combined context of what Cluely has read on your screen and heard in the meeting. Response quality depends on the LLM tier: the free plan uses a limited model; the Pro plan unlocks the latest available models. For technical interview questions, the response typically includes working code, explanation, and talking points. For behavioral questions, it provides structured talking-point responses based on common frameworks.
Stealth Overlay (Pro + Undetectability)
The stealth overlay, available on the $75 per month tier, is the product’s most technically novel feature. GPU-level rendering bypasses the capture layer that screen-sharing software reads, making the overlay invisible to remote participants viewing your shared screen. This is the feature that generated the most technical debate and the most controversy. It functions as advertised against standard screen-capture in Zoom, Meet, Teams, and Webex, but OS-level process detection tools can identify Cluely’s presence separately from screen capture.
Meeting Notes and Transcription
Cluely transcribes meetings in 12+ languages and generates structured notes from the session. This is the most straightforwardly useful feature for professionals who want AI assistance in meetings without the ethical complexity of real-time answer generation. The transcription and notes function independently of whether you use the AI response features during the call.
Multimodal Context
The combination of screen OCR and audio transcription means Cluely has broader context than a tool using only one input. If the interviewer shares a problem statement on screen, Cluely reads it. If they explain something verbally, Cluely hears it. The combined input produces suggestions that are more contextually accurate than audio-only or screen-only approaches.
Cross-Platform Support
Cluely works across Zoom, Google Meet, Microsoft Teams, Webex, and Slack huddles without requiring the other participants to install anything or the session organizer to enable any integration. It runs locally on your machine and never appears in the meeting interface that other participants see.
Cluely Pricing
| Plan | Price | Key Features |
|---|---|---|
| Starter (Free) | $0 | 5 AI responses per day, 100-character output limit, basic transcription |
| Pro | ~$20/month | Unlimited AI responses, latest LLMs, meeting notes, standard overlay (visible in screen share) |
| Pro + Undetectability | ~$75/month | Everything in Pro plus GPU-level stealth overlay invisible in Zoom, Meet, Teams, Webex screen share |
| Enterprise | ~$200/month | SSO, RBAC, post-call analytics, custom integrations, team management |
The free tier is genuinely limited: five responses per day with 100-character output caps makes it useful for evaluation but not for real use. The Pro plan at approximately $20 per month is the practical entry point for anyone seriously evaluating the tool. The Pro plus Undetectability tier at $75 per month is where the product’s flagship stealth capability lives, and it is priced for users who specifically need the overlay to be invisible during screen sharing rather than just visible to themselves.
The Data Breach: What Happened
In mid-2025, security researchers identified that Cluely had left admin credentials in a public GitHub repository. The exposed credentials allowed access to user data, resulting in the exposure of information on over 83,000 Cluely users, including payment-associated account details. Lee acknowledged the breach publicly.
For a product whose value proposition involves sitting in on every video call you make, access to your account data is particularly sensitive. The breach was the result of basic credential hygiene failure rather than a sophisticated attack, which is both more reassuring (it is fixable) and more concerning (it should not have happened). Anyone using Cluely should be aware of this history and make an informed decision about the trust they are placing in the platform’s data handling.
The ARR Controversy
In March 2026, Lee publicly retracted an ARR figure he had provided to TechCrunch in a 2025 interview, acknowledging that the $7 million number he had given was inaccurate. The retraction came via social media. The actual ARR figure was not disclosed in the retraction. This, combined with the data breach, gives potential users a relevant picture of the company’s track record on transparency during its growth phase.
Is Cluely Detectable?
The answer depends on the detection method being used and the plan tier. The stealth overlay on Pro plus Undetectability uses GPU-level rendering that makes it invisible to standard screen-capture software used by Zoom, Meet, Teams, and Webex. In that specific scenario, remote participants sharing your screen cannot see the Cluely overlay.
However, screen capture is not the only detection vector. Zero Assist, a detection tool designed specifically for this category of product, identifies Cluely’s presence at the OS process layer in under 500 milliseconds. This is separate from screen capture: it detects that Cluely is running on the machine, not that the overlay is visible. Dedicated detection products including Validia’s Truely and Proctaroo are in active development targeting the category of invisible AI assistance tools.
For most interview and meeting scenarios in 2026, Cluely is not detectable via standard video conferencing tools. For organizations actively deploying specialized detection software, it can be identified. As employers and testing platforms become more aware of these tools, detection capability will continue to develop. The stealth advantage that exists today is not permanent.
There is also a behavioral detection dimension that no technology addresses: AI-generated responses often have a recognizable structure and cadence that experienced interviewers may notice, even if they cannot prove why. The risk is not only technical.
Ethical Considerations
The ethical question Cluely raises is real and worth taking seriously rather than dismissing. Job interviews are supposed to assess a candidate’s own knowledge and judgment. If a candidate is using AI assistance invisibly, the employer is making a hiring decision based on a misrepresentation of what that person knows and can do unassisted. The friction this creates is not just philosophical: it affects who gets hired, what workplaces expect of employees once hired, and what happens when AI assistance is not available on the job.
The counterarguments are also real. Many interviews test how well someone performs under artificial high-stakes conditions rather than how well they will actually do the job. Professionals in most roles have access to search engines, documentation, and colleagues; testing people in isolation from those resources may measure the wrong thing. The “cheat on everything” framing is provocative, but the underlying product is a real-time AI assistant, a capability that exists in many forms across many legitimate professional tools.
Where individual use becomes clearest ethically is in contexts that explicitly prohibit outside assistance, academic exams and credentialing tests being the most obvious. Where it is most defensible is in general professional meetings, sales calls, and customer service interactions where real-time AI support tools are not prohibited and increasingly common.
Who Should Consider Cluely
Cluely is most defensible for professionals using it in contexts where real-time AI assistance is not prohibited and serves a legitimate function. Sales teams who want AI-powered call support, customer service representatives handling live queries, non-native speakers who benefit from real-time language support in professional meetings, and professionals who want structured meeting notes with AI context all have legitimate use cases that do not involve misrepresenting their capabilities.
For job interview use specifically, the risk profile includes not just detection but what happens after: if a company hires based on AI-assisted interview performance and the person cannot perform equivalently without that assistance on the job, the outcome is bad for both sides. The practical question is whether the interview format being used is actually measuring what matters for the role.
Anyone putting payment information into Cluely should be aware of the 2025 data breach and evaluate whether the platform’s current security posture is acceptable to them.
Verdict
Cluely works. The transcription is accurate, the real-time AI responses are contextually relevant, and the stealth overlay does what it claims against standard screen-capture tools. The controversy around the product was built intentionally and it succeeded at generating attention, but underneath the marketing is a functional real-time AI meeting assistant.
The honest assessment is that the product’s value depends entirely on the use case and the user’s risk tolerance. For legitimate real-time meeting support in professional contexts where AI assistance is not prohibited, the Pro plan at $20 per month is reasonable. For interview cheating specifically, the risks (detection via OS-level tools, employer TOS violations, behavioral signals that AI-trained interviewers recognize, post-hire performance mismatches) are real and not addressed by the stealth overlay alone. The data breach history is a genuine concern that should factor into any decision to trust the platform with sensitive account information.
Cluely is the most visible product in a category that is growing regardless of whether any individual product leads it. Real-time AI assistance during conversations is technically achievable, in demand, and not going away. What Cluely did was launch that category loudly and controversially enough that everyone now knows it exists.
Frequently Asked Questions
What is Cluely and how does it work?
Cluely is a desktop overlay application for macOS and Windows that provides real-time AI-generated suggestions during live video calls. It reads your screen via OCR and listens to your microphone, then uses an LLM to generate contextual responses you see in an overlay window that other call participants cannot see through standard video conferencing tools. It works with Zoom, Google Meet, Microsoft Teams, Webex, and Slack huddles. Responses are triggered by hotkey rather than appearing automatically throughout the call.
Who founded Cluely and why is it controversial?
Cluely was founded by Chungin “Roy” Lee and Neel Shanmugam, who were suspended from Columbia University in March 2025 for building Interview Coder, an earlier tool that provided AI answers during coding interviews. Lee launched Cluely in April 2025 with explicit “cheat on everything” marketing, including a viral video showing him using the tool on a date to fake knowledge about art. The product generated 70,000 signups in its first week and raised $15 million from Andreessen Horowitz within five months. The controversy also includes a 2025 data breach affecting over 83,000 users and a retracted ARR claim in 2026.
Is Cluely detectable by Zoom or Google Meet?
The standard Pro plan overlay is visible in screen sharing: the same as any other window on your desktop. The Pro plus Undetectability tier at approximately $75 per month uses GPU-level rendering that makes the overlay invisible to standard screen-capture software in Zoom, Meet, Teams, and Webex. However, OS process-level detection tools can still identify that Cluely is running on the machine even when the overlay is hidden from screen capture. Specialized detection products targeting AI assistance tools are in active development and will expand detection coverage over time.
What does Cluely cost?
Cluely offers a free Starter tier with 5 AI responses per day and 100-character output limits. The Pro plan at approximately $20 per month provides unlimited AI responses and access to the latest LLMs, but the overlay is visible in screen sharing. The Pro plus Undetectability tier at approximately $75 per month adds the GPU-level stealth overlay that is invisible in standard screen-capture scenarios. An Enterprise plan at approximately $200 per month adds team management, SSO, RBAC, and analytics features.
Can Cluely be used for job interviews?
Cluely functions during job interviews in the same way it functions during any video call: it reads and listens to the session and provides AI-suggested responses via a hotkey-triggered overlay. Whether using it in a job interview is appropriate involves both practical and ethical considerations. Practically, OS-level detection tools can identify Cluely’s presence even when the overlay is visually hidden, and behavioral signals from AI-generated responses may be noticed by experienced interviewers. Ethically, using invisible AI assistance in a context where an employer is evaluating the candidate’s unassisted capabilities involves a misrepresentation of what the person knows and can do without assistance.
What happened with Cluely’s data breach?
In mid-2025, security researchers discovered that Cluely had left admin credentials in a public GitHub repository, exposing data on over 83,000 users including payment-associated account information. The breach resulted from basic credential management failure. Founder Roy Lee acknowledged the breach publicly. The incident is a significant data security consideration for a product that runs continuously during video calls and has access to sensitive professional conversation content.
Is Cluely available on mobile?
Cluely launched an iOS app in 2026, expanding beyond its original desktop-only macOS and Windows availability. The mobile app extends the real-time AI assistance capability to phone-based calls and meetings.
What are the best legitimate use cases for Cluely?
Cluely’s most ethically straightforward use cases are professional contexts where real-time AI assistance is not prohibited and serves a genuine purpose: sales calls where reps need product specifications and objection handling support in real time, customer service interactions where accurate information retrieval matters, professional meetings where non-native speakers benefit from language assistance, and any meeting scenario where structured real-time notes and AI context are valuable. The product functions as a general real-time AI meeting assistant regardless of the “cheat” framing that dominated its early marketing.
