What does it cost to attack your defence?
Intruso answers with a price in dollars, not with a block rate. If attacking comes cheap, someone is already doing it — and the percentage will not tell you.
What do you need Intruso for today?
Find out what it costs to attack me
We train attackers from scratch against a mechanism and tell you the price, in dollars, of beating it — and how much hardening it actually buys.
See a full measurementFind out whether hardening fixes it
There is a ceiling no challenge crosses, because past a certain point attackers stop solving and start buying. No data, no credentials, and nothing you type leaves your browser.
Assess my policyFind out when my defence expires
The cost of attacking falls every year, and not at a single rate. That is why your security has an expiry date — and the route that kills it may not be today's cheapest one. The page dates that day, with nothing leaving your browser.
Date the expiryPut Intruso to work measuring my system
Six simple questions. The page works out, right there, what one solve would have to cost the attacker — and if the numbers come out small, it tells you that hiring me is not worth it.
Request a measurementGovern AI agents instead of blocking them
Legitimate automated agents are already customers. Instead of asking them whether they are human, give them a signed mandate — with scope, rate, expiry and tariff. Issue one here and try to break it.
Issue a passport nowIntruso runs exclusively against mechanisms that are yours, or recreated in a lab from your description. Never against third-party services: this is an audit, not an attack.
On an ordinary text CAPTCHA, attacking returns 769 times what it costs
Each attempt to beat it costs the attacker $0.000026. Each time one gets through, it is worth $0.02. The arithmetic is that division, and anyone can redo it.
Blocking 99.4% of attempts makes the attack 167 times dearer — and it still pays off 4.6 times over. That is why the block rate, on its own, does not tell you whether you are protected.
The question “is this a human?” has lost its answer
An AI agent buying on behalf of one of your customers is automated traffic and revenue. Blocking it costs you money. Letting it in with no rule at all costs you again.
The question that still has an answer is a different one: who answers for this, and how much do they pay? A signed mandate — who they are, what they may do, at what rate, until when, and at what tariff. Not blocking agents: charging them.
The question “who answers for this, and how much do they pay?” has one.
And there is a measured reason why this is not a fashion. Every defence built on a puzzle has an expiry date, because the compute that attacks it gets cheaper every year. Forging accountable identity does not ride Moore's law. The Horizon dates that expiry — and shows that the passport is the only defence outside the clock.
This has stopped being a sketch. The protocol is implemented and runs entirely in your browser: the key is born there, the mandate is signed with real Ed25519 and the verifier walks all nine points. The same logic exists in Python, in the engine, and the two are compared case by case by a harness anyone can run.
What is still missing is not code: it is the first production site serving requests with this. That is why the card says no customer yet and not product — and it will go on saying so until there is one.
Issue a passport now I want to be the first to put it into service
The cheapest way to start is without talking to us
The ceiling instrument runs entirely in your browser and makes not a single network request. If the result bothers you, then we talk.