Investigation units & quotas
logcat.ai meters AI usage on a single axis: the investigation unit. Your account gets a budget for its period, and every operation that runs AI over your data draws down against it. A quick look and a deep investigation are not the same amount of work, and a comparison across more files draws down more than a single-file look.
Your billing page in the console carries the numbers for your account: what each operation drew down, what you’ve used, what you have, and your reset date.
What each operation draws down
Every operation is priced on how much of your capture was analyzed. The anchor is simple: a Deep Research over 100 MB draws down one unit, and every other operation is a share of that rate over the same data. A megabyte here is 1,000,000 bytes. Before you start a quick search or a deep research, the console shows what it will cost for that capture.
| Operation | Rate | Minimum |
|---|---|---|
| Analysis on upload | 5 units per GB | 0.25 |
| Quick search | 1.5 units per GB | 0.10 |
| Deep research | 10 units per GB | 1.00 |
| Follow-up question | 1.5 units per GB | 0.10 |
| Comparison | 10 units per GB of everything compared | 1.00 |
| Source enrichment | Not metered | None |
The minimum is what a small file costs. An 8 MB kernel log draws 0.25 for its upload analysis and 1.00 for a deep research, because the rate over that little data would come to less.
Some worked examples:
| What you do | Size | Units |
|---|---|---|
| Upload a logcat | 60 MB | 0.30 |
| Upload a bugreport | 120 MB | 0.60 |
| Upload a baseband capture | 6 GB | 30.00 |
| Quick search on that bugreport | 120 MB | 0.18 |
| Deep research on that bugreport | 120 MB | 1.20 |
| Compare three bugreports | 360 MB combined | 3.60 |
A follow-up question is charged at the quick-search rate, and a conversation is capped. Once the follow-ups on one investigation have drawn down what a single deep research over the same data would have cost, the rest of that conversation is free. Asking again never costs more than starting over.
A comparison is priced on the combined size of everything it covers, so comparing more files, or larger ones, draws down more.
For the allowance that applies to your account, email sales@logcat.ai.
What is and isn’t charged
An operation that never ran is not charged. If a comparison is declined, or an investigation is stopped or fails before it starts work, nothing is drawn down for it.
A search that ran is charged, including one that found nothing. Searching is the work: a question that turns out to have no answer in your data took the same effort to answer as one that did. That also means a run that reaches your data and then fails partway can still draw down, because the work was done.
A search that could not read your data because a capture was still being prepared for querying is not charged: a quick search answers 503 and asks you to try again. A deep research first waits a few minutes for the capture to be ready; if it is still not ready, the investigation does not start and is not charged. A deep research that finds a capture being prepared partway through and reads nothing else is given its units back.
A deep research whose report could not be written gives you the findings it established with no conclusion. It is given its units back, and when you open the report it tells you it was not charged. A shared link, a PDF or an export of the report says only that it has findings but no conclusion.
The analysis that runs automatically when you upload a file draws down units too, in proportion to how much of the file was analyzed. An upload is one charge: a bugreport, an archive or a bundle is billed once for everything analyzed inside it, not once per file it contains. An archive costs at least 0.05 units for each file it holds to analyze whose analysis delivers, counted the way the archive limit counts them; a file that counts as several analyses counts for each one that was made, once any of them completes, and a file your plan doesn’t include analysis of isn’t counted. So 200 small logs that all complete draw down 10.00 even when their size alone would come to less. That minimum, counted on every file the archive holds to analyze, is checked against your balance before anything in it is analyzed; if it doesn’t fit, the upload is refused with the price, and nothing is drawn down. If an archive does not finish unpacking after a bugreport inside it has started analyzing, that bugreport is billed on its own, as if you had uploaded it alone.
The reset period
If you have a paid subscription, your budget resets on your renewal date each month. If you don’t, it resets with the calendar month. On a trial, you get one fixed 14-day budget that doesn’t refill during the trial. On an enterprise contract with a fixed term, your organization shares one pool for the whole term, and it does not reset each month.
Your billing page shows your current usage and the exact reset date for your account.
Running out mid-period
If you run out before your reset date, email sales@logcat.ai and we’ll raise your allowance. The new figure applies from the moment it’s set, so you don’t wait for the next period to use it. Storage works the same way. See Plans & pricing.
Next
- Why was I limited?: what happens when you hit the cap.
- Plans & pricing: capabilities and allowances.