Walking in is easy. Reaching anything is the nightmare.

Access is not reach.

For the intruder, every door opens onto an empty room. For you, it's quiet. NIGHTMARE.sh builds and tests your SaaS so that access is not reach, and shows you where it still is.

6 questions. We never ask for an IP, a host, a token or a key.

Scroll to follow the crossing

Access is not reach.

We learned this defending real products.

NIGHTMARE.sh — engraved prison medallion

NIGHTMARE.sh began as the defense of a SaaS in production. It took an automated attack against its public entrances. The same kind of entrance almost every modern product has: an API, a login, a form.

The attacker got in. That part was never in doubt, because those doors are public by design.

What mattered was what the doors opened onto. The answers came back "OK", and they came back empty. We saw no sign that any protected data was read. The structure held.

We also have to tell you the other half. Our emergency response was improvised, and it cost us availability. Our own containment took part of our own service down for a while. The attack did not do that. Our panic did.

We took two lessons from the same night.

One: build so that getting in leads nowhere. Two: never improvise the response. Rehearse it, before the bad night.

No names, no dates, no numbers on this page. This is what happened, in the shape we are willing to tell it.

Read the doctrine ↓

The door was never the problem. What it reaches is.

Access is not reach.

Most security is a wall: one line, one bet that nothing gets past it. But much of what you run is open on purpose. A public key in the front end. An API that generates itself. A login by link. When something walks through, you are blind, and you improvise.

The NIGHTMARE.sh Doctrine starts from the other end. Assume the entrance will be easy. Design what comes after.

Access is not reach.

1. Empty inside.

Even through an open door, anyone who shouldn't be there reads nothing. The room answers, and the room is empty.

2. Visible at every step.

The deeper someone goes, the more you know about them. Each step inward leaves a trace you can read.

3. Closable without bringing the house down.

Every route can be shut, and the response is rehearsed in advance. It is never invented during the incident.

The Doctrine is a way of building and responding. It is not a product you install, and it does not promise that nothing will ever happen. It promises that we know what we are doing when it does.

Find out where your system stands → Start the diagnostic

6 fronts. One question each.

Everything we do answers a question you already have. Here are the 6.

Access is not reach.

Data & Identity

Who can read what, and how would we know if they shouldn't?

Edge & Perimeter

What can an outsider reach right now?

Servers & Containers

If one machine falls, what else falls with it?

Application & Code

Where can our own code be turned against us?

Observability & Alerting

Would we notice? And would the right person hear it?

Incident Response

Who acts when an incident starts?

Weak at one front, and the others carry the load. Solid at all 6, and an intruder has nowhere to stand.

Answer the 6 → Start the diagnostic

6 questions. A plain answer.

We never ask for an IP address, a hostname, a token, a password or a key. Answer from what you know. If you don't know, say so: that is an answer too.

For each question, choose Yes, Partly, No or I don't know.

1. Data & IdentityCan you say, for each table or bucket that holds customer data, who is allowed to read it and who is allowed to change it, and have you checked that rule from the outside, not just read it in the code?
2. Edge & PerimeterDo you have a current list of everything on the internet that belongs to you (domains, subdomains, public APIs, admin panels), and is each one there on purpose?
3. Servers & ContainersIf one of your servers or containers were taken over today, could it reach the others, or your secrets, from inside?
4. Application & CodeHas anyone tried to break your application the way an attacker would (forged requests, other users' records, public forms), instead of only reviewing the code?
5. Observability & AlertingIf someone read data they shouldn't, or sent thousands of strange requests to a public endpoint, would an alert reach a person, and would that person know it matters?
6. Incident ResponseIf an attack started tonight at 22:00, is there a named person, a written first step, and a way to close a route without taking your own service down?

3 ways to work with us.

You do it. You do it with us. We do it for you. Choose how far you want us to go.

Free

Role

You do it.

Who it's for

Developers and technical founders building a SaaS, alone or in a small team, who want to know in an afternoon whether there's an obvious hole.

What you get

3 skills that run with your AI coding agent on your own project: pentest, hardening and incident response. Each finding comes with the evidence, the risk, a fix and a retest. Updates when the skills change.

What we need from you

Name, email and WhatsApp. A project that is yours to test.

How it starts

Fill the form and download right away.

Price

US$ 0. 3 fields instead of money.

Takes about two minutes.

Your system, your terms. We never see your project.

Assessment

Role

You do it with us.

Who it's for

CTOs and tech leads of SaaS companies already billing customers, with a team to fix things but no security specialist, and often a date: a big customer's questionnaire, an investor, a launch.

What you get

Written scope and a delivery date, before you pay. A scan across the 6 fronts. A report that says what to fix first, and why. Guidance while your team fixes. One retest of the findings you fixed.

What we need from you

Written authorization. A test environment you control. Test accounts you create and revoke. One contact on your team. Optional: read access to the code.

How it starts

You request, we agree scope, you get the delivery date in writing, then you pay.

Price

US$ 10,000, one time.

Request an Assessment

One retest included. Delivery date in writing before you pay.

We work in a test environment. We never ask for production access, passwords, keys or server addresses.

24/7 Protection

Role

We do it for you.

Who it's for

Founders and CEOs of SaaS companies where an hour of downtime or a leak costs more than they want to find out, and nobody is in charge of security.

What you get

The Doctrine applied to your system and kept up as it changes. A written, rehearsed incident response. Response terms in your contract: channel, hours covered, time to first action. The right to display Protected by NIGHTMARE.sh.

What we need from you

An application and a conversation. Time to map the system, write the runbook and rehearse it with us.

How it starts

You apply. We talk. If we fit, we deploy and rehearse. Coverage starts after the rehearsal.

Price

US$ 80,000 / month (about US$ 1M a year). The monthly fee starts when coverage starts.

Apply

By application · Coverage starts after onboarding

We accept a limited number of Protection clients, and we cover only what we have rehearsed.

Not sure which? Take the diagnostic. It recommends one plan, and shows you the rule it used.

Straight answers.

Free

Why do I have to sign up for something free?

Because 3 fields is the price, and we would rather say so. In return you get the whole kit and a notice when it's updated. You see what's inside before you fill anything in.

Why do you want my WhatsApp?

Two reasons. We message you when the skills change (a new probe, a correction). And one person from NIGHTMARE.sh will message you once, to ask how it went. That's all. Reply "stop" and it stops.

Will you spam me or sell my data?

No. We don't sell or share your data. You can leave in one step and ask us to delete it. Our privacy policy is next to the button, not buried in the footer.

Is it going to run something dangerous on my project?

The skills require an authorized environment, low request rates and stopping at the first sign of harm. Those rules are on the page before the form.

Isn't this just another checklist?

A checklist is a list. These skills follow the sequence of a real attack and hand you evidence, a fix and a retest for each finding. The page shows an example, marked as fictional.

What's the catch?

This is the door in. You get a real tool. We get to talk to people who ran it. When your list of findings outgrows your team's capacity to order it, there's the Assessment.

Will it work with my stack and my agent?

It was built for SaaS products with a database behind an API, an edge or proxy, a server and containers. It does not cover every stack. The install section lists the tools we have actually tested.

Assessment

US$ 10,000 for a PDF?

You're not buying a PDF. You're buying the order (what to fix first), the reason, the guidance while your team fixes, and a retest that confirms the fix worked.

A scanner is cheaper.

A scanner lists. It doesn't know what an attacker would do first in your project, it doesn't sit with your team, and it doesn't confirm the fix.

Do you need access to everything?

No. We work in a test environment that you control, with test accounts you create and revoke, under written authorization. Production only if you ask and it's written into the scope. We never ask for passwords, keys or server addresses through this site, by email or in chat.

I have a deadline. Will you make it?

You get the delivery date in writing, before you pay. If we can't meet the date you need, we'll tell you before you pay.

What if my team's fix doesn't work?

One retest of the findings you fixed is included. You get a record of what we verified as fixed. New findings from the retest go in the record. They don't open another included round.

My team has no time to fix things.

That's why the report is ordered: item one matters most. If there's truly no time and no people, that's a sign of 24/7 Protection.

You list no clients. Why trust you?

We don't list clients and we don't display badges we haven't earned. What we offer instead is the origin, the method, and the free skills, which you can judge before paying us anything.

What if you find nothing?

Then the report says what is solid, and why. That is also what you show a big customer or an investor. We won't promise a number of findings.

24/7 Protection

US$ 80,000 a month is a lot.

It is. It's for companies where an hour of downtime or a leak costs more than the year. If that isn't you, the Assessment is the right plan, and we'll say so. The monthly fee starts when coverage starts.

Why apply? Why can't I just subscribe?

Because we accept a limited number of clients and cover only what we've rehearsed on your system. Anyone who sells this in one click hasn't rehearsed anything.

Why doesn't coverage start when I sign?

Because improvised response is expensive, and we know that from our own history. First we map your system, apply the Doctrine and write the runbook with you. Then we rehearse it. Then we cover.

What does "immediate" mean? How many minutes?

Channel, hours covered and time to first action are written into your contract before coverage starts. We don't put on this page a number that isn't in your contract.

Will I be locked in?

Minimum term and notice period are written and shown to you before you sign. No hidden renewal, no surprise penalty. If you leave, you keep everything we built for you: the runbook, the configuration and the documentation. The badge leaves with the protection.

I already have my cloud and CDN protections.

Tools block patterns. They don't respond to incidents. Someone has to decide what to close without taking the service down, and that someone needs a runbook they've rehearsed.

What if your defense takes me down?

That happened to us, once, in our own origin story. It's the reason nothing is covered before it is rehearsed.

Who are you? How many of you?

A small, independent firm that takes few clients on purpose. We don't publish a team size, and we won't use words for staff we can't show.

Leave your email.

Interested in NIGHTMARE.sh? Leave a way for us to reach you.

Only your email and language are requested. No system details.

Find out what your door reaches.

6 questions. A plain answer. No IP, no host, no token, no key.

Start the diagnostic