In mid-2025, Cluely had a data breach that exposed personal data for 83,000 users. That’s not a minor footnote. If you’re choosing an AI assistant that runs on your computer during job interviews or sales calls, the security posture of the company behind it is a real consideration, not just a feature checkbox.
I want to be fair here. Data breaches happen to companies of all sizes. One incident doesn’t permanently define a product. But it does raise questions worth asking about infrastructure maturity, and it happened at a point when Cluely was also making performance claims that didn’t match what independent testers were finding.
What these tools are trying to do
Both Craqly and Cluely are real-time AI assistants that run during live calls or interviews. The category is sometimes called “AI interview copilot” or “live meeting assistant.” The core idea is that the software can hear what’s being said and surface relevant context, suggested answers, or supporting information on a private overlay visible only to you.
Cluely launched with significant attention, partly because of its founder’s backstory and partly because the demo was compelling. It positioned itself as a universal assistant for interviews, meetings, and general computer use. Craqly’s positioning is narrower: interviews, sales calls, and meetings within a workspace context, with a set of integrated products rather than a single general-purpose overlay.
The performance gap between marketing and reality
Cluely’s marketing materials claim a 300-millisecond response time. Independent testing found actual response times of 5 to 10 seconds in live conditions. That gap is significant. A 5-second lag during a behavioral interview question, when you’re trying to decide how to frame an answer in real time, is disorienting. A 300-millisecond assist feels invisible. A 5-second assist feels like you’re waiting for a tool to catch up.
Craqly tested at under 3 seconds in similar conditions. Not 300ms, but meaningfully faster in the scenarios where response time matters most.
I don’t know whether the 300ms figure was an early benchmark that didn’t hold under load, or aspirational marketing, or something in between. What I know is that the discrepancy is large enough to affect actual use.
Pricing and what you’re paying for
Cluely offers a $20/month Pro tier and a $75/month tier that includes the “undetectability” feature, software designed to make the AI assistance invisible to proctoring or detection systems used by some employers and interview platforms.
There’s a legitimate debate about whether that’s an ethical product to build or buy. I’ll leave the ethics question to each user. The practical question is whether $75/month is sensible math. If a job search takes 2-4 months and you’re spending $150-$300 on an interview tool, that math works if the tool meaningfully improves your outcome. Given the performance gap between marketing claims and real-world testing, I’d want to validate the actual response quality on my machine before committing.
Craqly’s paid tier runs around $15/month with a limited free option. No undetectability feature, and no equivalent pricing tier for it.
The founder controversy and what it signals
Cluely’s founder, Roy Lee, was publicly expelled from Columbia University and has been open about that story as part of the company’s marketing narrative. Depending on your perspective, that’s either a compelling “rules are made to be broken” founder origin story or a signal worth weighing when evaluating a company you’re trusting with data from your job interviews and professional calls. I think both interpretations are reasonable. The signal I find more useful is product execution quality, and on that front the performance data and stability reports are mixed.
Cluely experienced two complete freezes during one week of testing. Craqly remained stable in comparable testing. Software freezes during a live interview are high-cost events, you can’t restart an app mid-case-interview. Stability matters more for this category than for most tools.
Product scope: how broad is too broad?
Cluely has expanded toward being a general-purpose AI overlay for any screen activity. Craqly stays focused on professional call and interview contexts, with 8 integrated products across that domain (interview prep, mock interviews, meeting notes, sales call assist, and others).
Whether broader scope is better depends on what you’re buying the tool for. If you want a single AI assistant for everything on your screen, Cluely’s ambition fits. If you want a tool specifically tuned to performing well in high-stakes professional conversations, focused tooling tends to do that better. The 2024 Stack Overflow Developer Survey found that developers strongly prefer specialized tools for specific tasks over general-purpose AI assistants for performance-critical work. That finding probably generalizes beyond development.
Who should use which
Cluely may make sense for users who specifically want the undetectability feature and have evaluated the ethical tradeoffs, or who want a more general-purpose screen overlay for diverse tasks. If the company addresses the stability and performance gap issues, the product experience could improve significantly from where it was in mid-2025.
Craqly makes more sense for users who want reliable, lower-latency assistance specifically for interviews and sales calls, at a lower price point, from a company that hasn’t had a major data incident. The LinkedIn Economic Graph’s research on interview preparation trends suggests that AI-assisted interview prep is becoming normalized across job functions, which means the stakes of picking a stable, trustworthy tool are higher than they were even a year ago.
Neither tool is obviously the right answer for everyone. The breach history, the performance gap, and the price difference are all real inputs to a decision that depends on your specific situation and risk tolerance.