Why a deny costs the same as an allow

The most common question about our pricing is some version of "wait, I pay for a no?"

Yes. A deny costs exactly what an allow costs, and we did that on purpose. It comes straight out of what we actually sell.

What you're paying for#

When a business asks Delegus whether an agent may do something, it gets an answer back, along with a signed receipt of that answer. The receipt records which company allowed which agent, under which permission, and what was checked. When the answer is no, it also names the first check that failed. Anyone holding it can re-check the decision later on their own machine.

So the receipt is really what you're buying. An allow is worth having because you can prove it later. A deny is worth more. It's the decision you're most likely to have to defend, like an order you turned away or money you didn't take. If a buyer ever says "you should have accepted that", or a regulator asks why something was refused, you hand them the deny receipt.

Charging less for a no would mean valuing the most important decisions least.

We're not paid to say yes#

There's a second reason, and it's about incentives.

Imagine a check that earned more for saying yes than for saying no. The service answering would have a reason to lean towards yes, and we don't want that reason to exist, even in principle. A check costs the same whatever the answer. The only thing the answer depends on is the company's permission, evaluated the same way every time.

That matters most to the business relying on it. You're trusting the check to protect you, so it shouldn't have any stake in how it comes out.

The seller pays; the company issuing permissions doesn't#

The business that runs the check pays for it. That's the seller, or whatever service is about to act; our spec calls it the relying party. A company that proves it owns its domain and signs permissions for its agents is never charged.

We set it up that way on purpose. A permission only helps if the businesses its agent deals with can check it, and checking only helps if companies bother to write permissions in the first place. Charging the companies that write them would slow down the very thing that makes checking worth doing. So whoever benefits at the moment of the decision pays, and whoever makes the decision possible doesn't.

There are no seats and no per-user fees. You pay for checks.

What that looks like in practice#

• Sandbox: free, no card, with test permissions only. Wire up the check and see a signed receipt.
• Going live: a 14-day trial when you add a card, then Starter, $99 a month with 50,000 checks included.
• Growth: $499 a month with 100,000. Scale: $1,500 a month with 1,000,000. Overage per check steps down as you move up: $0.030, $0.015, $0.008.
• Founding pilot: $1,500 a month, flat, for up to 2 million checks ($0.004 above), with the price locked for 12 months and 30 days' notice to stop.

In every tier, an allow and a deny are the same line on the bill.

One run, on a bill#

The recorded run on our homepage makes four decisions. Acme's agent orders $18,400 of compute from ComputeCo, and that one is allowed. Then it tries $31,000, over its $25,000 limit, and is refused. It sends the same order to a seller Acme never named, and is refused again. After Acme takes the permission back, the original order is refused too. Run through our service, that's four checks on the bill, and every one costs the same.

Now ask which of those four a business would most want a signed record of. The allow let through an order that was always going to be fine. The three denies are where ComputeCo turned money away, or stopped an agent its buyer had already switched off. Those are the receipts that earn their keep.

See pricing.