Private access · AI agent defence

AI agents are probing your systems. Gateway shuts them out.

Autonomous AI attackers scan, guess and break in around the clock, with no person at the keyboard. Gateway picks them out from your real customers and cuts off every way back in: their address, their sessions and any credentials they took.

Python, Node.js or nginx Installs in minutes Private access, by approval
The threat

Attackers no longer need to be awake.

An AI agent can be handed a single instruction, "find a way in", and left to work through your site, your API and your sign-in pages for as long as it takes. Firewalls and rate limits were built for people and simple scripts. Agents read, adapt and try again.

What Gateway does

Identify. Kill. Revoke.

Gateway runs alongside your application and judges every request. When it is sure it is looking at an attacker, it does not just block one request. It ends the attacker's access everywhere.

01

Identify

Tells AI agents and attack tools apart from your customers, your partners' integrations and honest search crawlers. Every decision comes with plain-English evidence your team can read.

02

Kill

One decision shuts the attacker out of every site in your account at once, and keeps recognising it when it comes back from a different address.

03

Revoke

A signed notice tells your own systems exactly which sessions and API keys the attacker used, so you can cancel them. The agent loses access inside your app, not only at the front door.

A console your team will actually use

See who came, what they tried, and why each one was stopped. Lift any decision in one click. Every change is recorded in an audit log.

Stronger together

When one member proves an address is an attacker, every member's protection hears about it within seconds. Shared intelligence raises suspicion; it never blocks your visitors on its own.

Reliability

Built so it can never become your outage.

A security tool that takes your site down has done the attacker's job for them. Gateway is designed around that rule.

300 ms

Your site never waits

If Gateway has not answered in 300 milliseconds, the request goes through. You can choose to refuse instead on your most sensitive pages.

0

Outages do not reopen the door

Your installation keeps its own copy of who has been shut out, so attackers stay blocked even if Gateway itself is unreachable.

1 click

Every decision can be undone

Lift a block and clear that visitor's record in one step. Add your offices, monitors and partners to an allowlist that is never touched.

Safe

Starts cautious

New sites act only when Gateway has proof, never on a hunch. Review what it would have done for as long as you like, then turn on full protection.

Your users' secrets stay on your servers

Passwords, cookies and API keys are replaced with irreversible codes inside your own server before anything is sent. Gateway never receives them in readable form.

Install

Minutes, not a project.

Once approved, create a site in the console, copy your key, and add one of three small integrations. No change to your hosting, your DNS or your code paths.

Python: Flask, Django, any WSGI appNode.js: Express and Connectnginx: any backend language
# Python from gateway_shield import GatewayShield app.wsgi_app = GatewayShield(app.wsgi_app, key=os.environ["GATEWAY_KEY"], endpoint="https://gateway.dendat.ai") // Node.js app.use(require("./gateway-shield")({ key: process.env.GATEWAY_KEY, endpoint: "https://gateway.dendat.ai" }));
Who it is for

Any US business with something worth breaking into.

If you run a customer portal, an online store, an API or an admin panel on the open internet, AI agents can reach it.

SaaS platformsOnline storesBanks and fintechHealthcare portalsInsuranceLaw and accounting firmsGovernment contractorsSchools and universitiesLogisticsAgencies hosting client sites
FAQ

Questions teams ask first

Will it block my real customers?

New sites only act when Gateway has proof that a visitor is an attacker, and real customers do not produce that proof. Everything else is recorded as "would have blocked" so you can review it before you turn on full protection. Any decision can be lifted in one click.

Will it slow my site down?

Your site never waits more than 300 milliseconds for an answer. If Gateway is late, the request goes through.

What happens if Gateway goes down?

Your site keeps working, and attackers that were already shut out stay shut out from a copy kept on your own server.

Do you see our users' passwords or cookies?

No. They are replaced with irreversible codes on your server before anything leaves it.

Will search engines still find us?

Yes. Honest crawlers that identify themselves are allowed by default. You can choose to keep AI training crawlers out as well.

How does detection work?

We share the details with account holders inside the console, not on public pages. The less an attacker can read about how it is caught, the more reliably it is caught.

How do we get access?

Gateway is private. Request access with a few details about your company and what you want to protect. We review every application by hand and onboard a limited number of companies at a time, so each one gets direct help with setup.

What does it cost?

Pricing is agreed with each company during onboarding, based on the sites and traffic you protect. Cancel anytime.

Request private access to Gateway.

We onboard a limited number of companies at a time. Tell us what you want to protect and we will be in touch.

Request access