Almost every AI interview copilot has a free tier, and in most of them the free tier excludes the one feature people actually came for. Not by accident, and not really out of greed either. There’s a structural reason, and once you see it the whole category’s pricing makes more sense.
The short version: staying hidden during a screen share requires a native desktop application, native applications are expensive to build, sign, notarize and distribute, and so that capability tends to land behind the paywall while the free tier gets a browser extension that structurally cannot do it.
Why stealth can’t be free in a browser tool
Being invisible on a screen share is not a clever trick. It’s an operating system call.
Windows exposes SetWindowDisplayAffinity with a WDA_EXCLUDEFROMCAPTURE flag that tells the compositor to draw a window on the physical display but leave it out of any capture stream. macOS has an equivalent in NSWindow.sharingType. Desktop frameworks wrap both, which is what Electron’s setContentProtection(true) does.
A Chrome extension has no access to any of that. It isn’t a window in the OS sense, it’s content inside Chrome’s window, and Chrome is the thing being shared. There’s no setting, no upgrade, no configuration that grants an extension this capability. It’s an architectural boundary.
So when a product’s free plan is “browser extension” and its paid plan is “desktop app”, the free plan isn’t a smaller version of the paid one. It’s a different thing that cannot do the primary job. Worth checking before you spend a week evaluating it.
The three shapes of “free” in this category
They get advertised with the same word and behave nothing alike.
- Free trial. Full features, fixed window, card usually required up front. You get a real evaluation, and you have to remember to cancel. Fine if you’re interviewing next week, useless if your process runs two months.
- Freemium with features removed. Permanent, no card, and the removed feature is very often the one that matters. This is where the browser-extension-versus-desktop split usually shows up.
- Credits or minutes. Permanent, no card, full features, capped usage. You can verify the thing works on your own hardware, you just can’t lean on it every day.
The third shape is the only one that answers the question most people are actually asking, which is “does this work on my machine, in my setup, before I pay anything”. A trial answers it with a deadline attached. A feature-stripped freemium answers a different question entirely.
What to check before you trust any free tier
Four things, in rough order of how often they surprise people:
- Is stealth included, or paywalled? Ask this first. If it’s paywalled, the free tier can’t tell you whether the product works for you, because the thing you need to verify is exactly the thing you can’t access.
- Desktop app or browser extension? This determines whether stealth is even possible, so it’s really the same question asked a different way.
- Does it join the call as a participant? Bot-based tools connect through a meeting API and show up in the participant list where everyone can see them. That’s fine for note-taking, fatal for an interview.
- Does the free usage reset? A one-time grant of credits and a monthly allowance are very different products wearing the same label.
I’d add a fifth that’s harder to check: whether the free tier is a real product decision or a lead magnet that gets discontinued. I don’t have a reliable way to predict that, and neither does anyone else.
What Craqly’s free tier actually is
Craqly’s Starter plan is $0 and includes stealth. It gives you 20 credits a month, where one credit is one minute of live session, and credits reset monthly. All five session modes are on it too. No card.
The reason stealth is on the free tier rather than behind the paywall is narrow and practical rather than generous. Capture exclusion is a request to the operating system, and the operating system’s answer varies by version and hardware. Apple has tightened screen recording permissions across several recent macOS releases and each change has broken window exclusion for some applications until they shipped a fix. Which means the only honest way to know whether it works for you is to test it on your machine.
Paywalling that would mean asking people to pay to find out whether the product functions on their hardware. Twenty minutes is enough to run the test properly and still have time for a short practice round.
Here’s the test, and it takes about six minutes:
- Start a solo meeting in Zoom, Google Meet, or Teams.
- Join again from your phone, or start a local recording.
- Open the assistant so it’s plainly visible on your display.
- Share your entire screen. Not one window, the whole screen, because that’s the harder case.
- Check the phone or replay the recording.
If the overlay appears there, exclusion isn’t working on your setup and no plan tier changes that. Retest after OS updates, since that’s when it quietly stops working.
Where a free tier genuinely runs out
Twenty minutes a month is a real constraint and I’d rather be straight about it. It covers the setup test plus a short screen. It does not cover a full loop, a week of sales calls, or daily standups.
For reference, and verified against the pricing page on 22 July 2026, the paid tiers are $19 a month for 180 credits, $39 for 420, and $69 for 780. Stealth is on all of them, since it’s already on the free one. Paying yearly takes 20% off. Usage is reserved when a session starts and refunded if unused, so a session you abandon after two minutes costs two minutes.
If you’re interviewing at one company, the free tier is probably enough. If you’re running a full search across six or seven processes, it isn’t, and I’d rather you know that going in than discover it mid-loop.
Start on the free tier and test it properly. 20 credits a month, stealth included, all five modes, no card required. Enough to verify it works on your hardware before you decide anything.
One thing worth saying about free tools generally
A free interview copilot doesn’t change whether using one undisclosed is acceptable to the company interviewing you. Most employers prohibit it, and a 2025 SHRM survey found a majority of hiring managers would disqualify a candidate over undisclosed AI use. Price has nothing to do with that.
The use that holds up under scrutiny is preparation. Running practice rounds with support until the patterns are genuinely yours, so that in the real interview you’re reasoning rather than reading. That’s slower and less satisfying than a panel feeding you answers, and it’s the version that survives a follow-up question.
Frequently asked questions
Is there a genuinely free AI interview copilot?
Yes, though “free” covers three different models. Free trials give full features for a fixed period, freemium plans give permanent access with features removed, and credit or minute-based plans give permanent access to all features with a usage cap. Only the third lets you verify the product works on your hardware without a deadline or a missing feature.
Why do free interview copilots usually lack stealth mode?
Because stealth requires a native desktop application. Hiding a window from a screen share means calling an operating system API such as SetWindowDisplayAffinity on Windows or NSWindow.sharingType on macOS, and browser extensions have no access to those. A free tier built as a Chrome extension cannot offer stealth at any price.
Does a free plan need a credit card?
It depends on the model. Free trials usually require a card and convert automatically. Freemium and credit-based free plans generally don’t, since there is nothing to convert. Craqly’s Starter plan requires no card.
How much interview time do 20 free credits actually cover?
Twenty minutes of live session, since one credit equals one minute. That covers setup testing plus a short screen. Credits are reserved at the start of a session and refunded if unused, so short sessions only consume what they use. It is not enough for a full interview loop across several companies.
Can I tell whether a tool joins my meeting as a participant?
Yes, and you should check before an interview. Start a solo meeting, connect the tool, and look at the participant list. Bot-based tools that connect through a meeting API appear there visibly. Desktop tools that capture system audio locally never join, so the list stays unchanged.