A monthly plan keeps the watch on. Everything beyond it is metered at published rates, every charge listed below, every charge listed below.
A wrong invoice is a refund. A wrong record is an afternoon. Those you can pay back. The customer who quietly decides you are careless is the one you never get back, and there is no amount of money that repairs a reputation once it goes.
Most companies don't realize they start every day behind the line, because the work happens where they can't watch it. AI is remarkable, and it is volatile. It is built to please and to finish. Finishing is the closest thing it has to a want, and it will finish without ever reconciling what it broke along the way.
Until AI governance catches up to what these systems are actually doing, checking them yourself is the fix. What that costs per month is just the cost of doing business. Every price is below, in full.
$99 / month
The basic watch. Enough to catch a problem early.
$299 / month
The plan we recommend first. a signed, defensible receipt arriving every month whether you remember to ask or not.
$799 / month
For shops where an automation quietly going wrong costs more than the plan does in a year.
Included quotas reset monthly and never roll over into a surprise. Metering only begins after your plan's quota is used, and you can set a hard monthly ceiling so a runaway system can never spend more than you allowed.
Or start without a plan$149 once
Available when our library already covers your workflow type. If yours needs a kit built from scratch, we'll say so and quote the Release Review instead, we don't stretch a fixed price to cover custom work.
$1,500 fixed
For workflows that touch serious money or big clients, and for any workflow we've never seen before.
Start with the free written fit check. If your workflow qualifies, you get a written scope, a plan recommendation, and a simple invoice. Payment happens only after both, we don't take money for work we can't verify honestly, and we decline jobs that don't fit rather than stretch them.
Passwords or logins · access to your systems · your customers' data · real payment records · regulated or personal information · your source files. Every test runs on fake data we invent, in your sandbox, by your hands. Because we never ask for access, hiring us doesn't require an IT review.
Not certification. Not a security or compliance audit. Not legal advice. Not monitoring. Not a guarantee your workflow is perfect. It is a limited, honest inspection with a signed verdict, and the receipt states exactly what was and wasn't tested, so it never claims more than it proves.
Self-serve, metered: load a balance and let your own systems pull a test kit (a few cents), score a run against the answer key (cents to a quarter), or request a full compound verdict, priced on the outcome, so you pay only when the answer is actually usable. Scheduled random drills included. In development now; founding clients get first access and founding rates locked.
Every rate on this page is published, fixed, and charged after a quota you already know. Nothing is estimated, nothing is billed by the hour, and the full verdict is charged only when it returns a usable answer, if the evidence can't support a conclusion, you don't pay for one. You can cap your monthly spend to the dollar.
Any change to your workflow makes the old receipt historical. New version, new integration, new AI model underneath, all of it counts. Re-check on meaningful changes or on a rhythm that matches how often their automations evolve.