BACKDOOR
They hack you for ransom. We hack you for the report. A swarm of AI agents attacks your site from the outside — exactly how real attackers would. No source code. No backend access. Just a real-world attack, followed by a report with proof, exploits, and fixes.
Autonomous AI for speed and continuous coverage. Human-led testing for depth and the assessment someone has to sign. Whatever you expose, we cover it — and we tell you which one you actually need.
A swarm of AI agents attacks your site from the outside, black-box, and proves every finding with a working exploit. Nothing to install, no access to your code or servers.
Our in-house testers go after business logic, access control between roles and the chains that turn three medium findings into one critical — then walk your team through the report.
Not sure which one? Start with the scan. If your case calls for people, we will say so.
AI penetration testing sends a swarm of AI agents to attack your systems the way a real adversary would: mapping your attack surface, hunting for weaknesses, and exploiting them to prove real impact. A scanner flags what might be a problem. AI pentesting proves what is — so your team fixes exploitable risk instead of chasing false positives.
Runs black-box, from the outside — nothing to install, no access to your code or servers.
Covers sites and web apps, domains and subdomains, servers and open ports in a single scan.
Proves every finding with a working PoC. Our rule: Found ≠ Vulnerable; Vulnerable = Exploitable.
A BackDoor security engineer reviews every report and screens out false positives.
Real attackers are already scanning your site. Let our agents go first.
Full report in 5 hours.
From € 5,100 · Black-box · Zero data retention · NDA available
| Vulnerability scanner | Manual pentest | BackDoor pentest |
|---|---|---|
| Time to result | ||
| Minutes, but shallow | 4–6 weeks | 5 hours |
| Proof of impact | ||
| None — "possible" issues | Yes, if time allows | Working PoC for every finding |
| Attack-surface coverage | ||
| Signature-based | Limited by hours | Sites, domains, servers in one pass |
| False positives | ||
| High | Low | Low — exploit-proven, then screened by a human |
| Frequency | ||
| On demand | Once or twice a year | On a schedule, as your perimeter changes |
BackDoor pairs the speed of automation with the depth of a real attacker — without the two-week wait or the false-positive noise.
Our default scan runs from the outside with no credentials — the same position a real attacker holds on day one. That is the right place to begin, and it is not the whole picture. Most of what a product actually protects sits behind the login.
The agents map your perimeter and attack everything exposed to the internet — no accounts, no inside knowledge, nothing installed.
Hand the agents a set of test accounts and the same swarm carries on past the login screen — into billing, admin settings and one tenant's data sitting next to another's.
Nothing to install, no access to your code or your servers, and credentials are used only inside the testing window you approve.
Talk to us about scopefrom € 10,000
from € 5,100
4–6 weeks
5 hours
2–4 weeks after first call
right now
Three steps. Same playbook a real attacker uses. Different ending.
We see what an attacker would see.
Forms, APIs, admin panels, subdomains.
~30 min
Every vulnerability, proven by exploit.
Steal data · bypass auth · reach payments · chain bugs
~4 hours
Built for your CEO and your engineers.
Working PoC · CVSSv3 score · step-by-step fixes
~30 min
Tell us about your scope or ask a question — we reply within one business day.
Every scan ends in a report your team can act on the same day — not a raw dump of alerts.
Every finding ranked by CVSS and sorted by priority, so you fix the riskiest issues first.
A working proof of concept for each vulnerability, so you see exactly how it is exploited.
Two reports in one: an executive summary for leadership and full technical detail for engineers.
Clear, step-by-step remediation guidance.
A free retest after you fix, to confirm the issue is closed.
Need audit-ready evidence? Map findings to SOC 2, ISO 27001, NIST, PCI DSS, and GDPR on request.
6 vulnerabilities found. 3 give direct access to customer data. Fix immediately.
Vulnerabilities found
2 critical, 2 high
POC with exploit
not "possibly vulnerable" — proven
Entry points tested
forms, APIs, admin panels, subdomains
Customer database access via contact form
Remote code execution through file upload
Auth bypass — access to other users' orders
Admin panel exposed without 2FA
Rate limiting missing on login — brute force possible
Want to see the actual report?
38 pages · proven exploits · CVSSv3 · step-by-step fixes
Pricing is shown per asset. The second and each additional asset costs $5,000/month.
$2,900 one time
For when you need to test a site once.
$6,000 month per asset
$6,000 billed monthly
Try it with no commitment. Cancel anytime.
$5,700 month per asset
$17,100 per quarter
Twice the compute power per target and automated patching.
$5,100 month per asset
$61,200 per year
Continuous mode: the agent never stops.
100+ companies tested.
127+ AI agents test in parallel across 300+ attack scenarios — so more of your perimeter is checked in a single pass.
Zero data retention — your data isn't kept after the engagement.
Safe for production: a gentle testing mode, a scan window you approve, and a stop button you control.
Authorized testing only — a scan starts after you verify you own the domain.
No install. No access to your servers. Authorized on your domain only, and you can stop anytime.
AI pentesting uses AI agents to actively exploit weaknesses and prove impact, not just list "possible" issues. A scanner matches signatures; BackDoor behaves like an attacker and validates each finding with a working PoC.
Yes. The agents find and exploit vulnerabilities on their own, and every finding has to be proven with a working PoC. A BackDoor security engineer then reviews every report and screens out false positives. The review is a filter, not a second manual pentest — if you need testers working through your application by hand, that is our manual penetration testing.
Yes. BackDoor runs in a gentle mode without load-style attacks, tests within a window you approve, and can be stopped at any moment.
A full external pentest completes in about 5 hours, versus the two-to-four weeks a manual engagement typically needs.
A scan only starts after you verify you own the domain. BackDoor tests your external perimeter with your permission — never anything you don't control.
Yes. The report is written in two layers — a one-page executive summary anyone can read, and a detailed engineering report with exact fix steps. Most teams hand the engineering part to their developers or hosting provider.
No. We run at a configurable rate limit and avoid destructive payloads by default. You can also schedule scans for off-peak hours.
Not for the basic scan, but NDA and DPA are available on request and required automatically for Scale and Enterprise plans.
Zero data retention by default. The report is delivered as a PDF and removed from our infrastructure after 30 days unless your plan requires longer storage.
Yes — our reports are mapped to SOC 2 CC controls, ISO 27001 Annex A, NIST CSF, GDPR Art. 32 and PCI DSS v4.0. Many of our customers use them for audit evidence.
Yes, an attestation letter is included with every passing re-scan. White-label and co-branded options available on Scale / Enterprise.
Coverage and consistency. Our agents test thousands of attack scenarios in parallel across every form, API, and subdomain — no human team can match that surface area at this price.
Yes — download a redacted real-world sample from the report card above.
We've never run a scan that found absolutely nothing — but if we genuinely return zero confirmed findings, we refund 50% and re-test in 30 days for free.
Both — start with one scan, upgrade to Pro, Scale or Enterprise whenever you need recurring coverage.
Real attackers are already scanning your site. Let our agents go first. Full report in 5 hours.
From € 5,100 · Black-box · Zero data retention · NDA available
