Upload a log export, get real totals for your whole network, free. This page states exactly what the preview shows, what it withholds and where the limits sit, so nothing surprises you.
These are complete, network-wide numbers, not samples. The preview totals are the same totals the paid report shows.
Every AI tool found, counted per functional category. The full distribution, nothing sampled.
How many high-risk tools, how many that train on user data, how many abusive-purpose tools. Real counts.
A fifth of the AI tools found, at least five, listed by name with category, risk and training verdict.
Every unnamed tool appears as a masked row in the table. You see exactly how many findings exist.
The preview renders as a PDF you can forward. Most upgrade decisions happen after someone shares it.
Run it monthly within the free caps. Rising category totals are a signal even without every name.
The preview does not hide that it is a preview. It shows you the shape of what you have not paid for.
...and one masked row for every further tool found.
A preview that silently truncates makes the network look cleaner than it is. That would be a dangerous document to circulate.
That last point is deliberate. Knowing that high-risk tools exist is free; knowing which ones costs $99.
The caps exist to keep the preview useful for evaluation while the full report carries the commercial load.
| Limit | Free preview | Why it is set there |
|---|---|---|
| Registration | Email address only | Reports need an account to live in for 90 days. |
| Runs | 1 a day, 3 a month | Enough for a baseline and a re-check, not for free monitoring. |
| Export size | 5,000 lines / 2 MB | Covers small networks fully and gives large ones an honest sample. |
| Tools named | A fifth, minimum five | The biggest habits become visible; the inventory stays paid. |
| Per-user table | Not included | Identity-level findings are the full report's core value. |
| CSV export | Not included | Machine-readable output ships with the paid tiers. |
| Sanctioned split | Counts only | The named split needs the full tool list by definition. |
Free or paid, the uploaded export is read once and discarded. Reports stay in your account for 90 days and can be deleted earlier.
Small networks often fit inside the free tier entirely. This is what that looks like.
Seven days of DNS logs from their filter: 3,800 lines, well under the 5,000-line cap. Identity column included.
14 AI tools across 6 categories. Two flagged as training by default, none abusive.
The top five by usage cover the design tools the founders already suspected, plus one transcription tool they did not.
Nine masked rows were enough to justify one $99 report. The full list took the Friday meeting from guesses to a policy.
At 40 people this network fit the free cap with room to spare. The sizing table shows where larger networks land.
We would rather you use the right tier than upsell you. The split is clean.
The upgrade is a one-time $99, not a subscription, and packs bring it to $60 a report. The math lives on the pricing page.
Before running anything, flip through the sample evidence pack. The preview you get free is this document with most rows masked.
You get one preview a day. Spend it like it costs something.
Blocked events measure your filter. Allowed events measure your exposure, and they use the 5,000 lines better.
Avoid holiday weeks and company offsites. One representative week beats a strange fortnight.
The preview does not show per-user data, but running the same export after an upgrade does. Same file, more report.
If your console can filter to web or DNS categories, do it. Mail and infrastructure noise spends lines for nothing.
The five summary tiles are the meeting-starter. They fit one slide and need no explanation.
Put next month's preview in the calendar before reading this month's. Trends beat snapshots, see the cadence loop.
The whole free path, from landing to PDF, with nothing skipped.
Email address, confirmation link, done. No company fields, no phone number, no qualification form.
Drag the export onto the panel. The parser announces the detected format and the line count before running.
Hostnames are extracted, deduplicated and matched. The upload itself is discarded the moment parsing ends.
Totals land first: tools found, high-risk count, training exposure, abusive tools, unsanctioned count.
The named fifth appears with verdicts, followed by one masked row per withheld tool.
Download the preview pack, forward it, and decide with your team whether the masked rows are worth $99.
Even without every name, complete category totals settle real arguments.
"Are we generating client-facing content with unvetted image tools?" A non-zero Image & Visual total answers it.
"Is company code leaving through code assistants?" The Code & Development count says whether the conversation is needed.
"Are customer conversations being pasted into chatbots?" Text & Language plus transcription totals frame the exposure.
"Is this a real problem or IT paranoia?" A tile row with red in it ends that debate in one glance.
"Do we have tools that train on data by default?" The training count is exact, even in the free tier.
"Anything abusive-purpose on our network?" That tile being zero is worth documenting; non-zero is worth escalating.
Department-by-department anatomy of where shadow AI concentrates lives on the examples page.
Security vendors use "free" three different ways. Only one of them respects your time.
| The pattern | What it actually costs | This preview |
|---|---|---|
| "Free assessment" = sales call | An hour with an SDR before any data is touched. | Self-serve. Nobody calls unless you ask. |
| "Free trial" = clock and card | 14 days, card up front, cancellation admin. | No card, no clock. Caps, not deadlines. |
| "Free scan" = fear report | Vague severity scores designed to alarm. | Exact counts, named top tools, and the gaps drawn honestly. |
The upgrade path is a $99 one-time purchase, not a subscription negotiation. That is the whole commercial model, detailed on the pricing page.
Yes. No card, no trial clock, no sales call. Registration needs an email address because reports live in an account.
The register behind the audit is a daily-maintained commercial product. The preview shows real totals; the full inventory is what you pay for.
Complete, across your whole export. Only the tool names are partially withheld, never the counts.
The fourth run in a month asks you to wait or upgrade. Caps reset monthly, and paid reports have no such cap.
Buying a report unlocks full runs. Re-upload the same export and the full report covers the identical window.
Shorten the window for the preview, or go straight to a full report which takes 25 MB and 2,000,000 lines.
Five reads, in order, and what each one decides. Ten minutes total.
Zero means proceed calmly. Non-zero means the full report, today, and probably HR in the loop.
Add "trains by default" and "opt-out" tools together. That number is how many vendors may already hold your prompts.
Write down the five tools you expected before opening the table. The overlap with the named fifth is usually under half.
Multiply by your discomfort. Twelve unknown tools in a regulated business reads differently than in a design studio.
Each category total should have an owner. A category with a big total and no obvious owner is your first meeting invite.
Enough signal? Book the monthly re-run. Not enough? One report, $99, and the decision-aid above tells you which you are.
Each cap has one job. None of them is "make the free tier annoying".
Daily classification of ~300,000 domains is a paid operation. Unlimited free inventories would be the product for free.
Per-day and per-month caps stop scripted walking of the register through repeated uploads.
Site-wide daily ceilings keep run times in minutes for everyone, free and paid alike.
Service providers run the free tier differently: not on their own network, but as the first ten minutes of a client conversation.
During a quarterly review, the MSP asks the client for a one-week DNS export. The client's own admin pulls it in five minutes, so no access questions arise.
The preview comes back with 21 tools across 8 categories, two of them training on data by default. The client expected "maybe ChatGPT".
The masked rows do the selling. Nineteen unnamed tools on the screen turn a routine QBR into a scoped assessment engagement.
The MSP delivers the full audit through their plan allowance, with the client's name on the report. The whole play is costed on the MSP page.
One log export, one upload, and the guessing stops. The preview tells you whether there is anything worth paying to see.
Run the free audit