pentest · offensive security
We test your URLs, applications, APIs and infrastructure the way an attacker would, with written authorization and a closed scope. AI agents widen the coverage; specialists validate every finding. You get what is exposed, the proof and how to fix it.
Each track can be hired on its own or combined. The scope comes out of the strategy call: what is critical, what can be touched and in which window.
We map everything that answers on the internet under your company name: domains, subdomains, routes, admin panels and files nobody remembered were public. Then we check every address.
Controlled exploitation of web applications, APIs and infrastructure, in black, gray or white box. The focus is whatever lets someone get in, escalate access or see another customer’s data.
A review of the configuration around the code: cloud, secrets, dependencies, deploy pipeline and personal data handling. This is where a large share of incidents start.
AI agents run reconnaissance, generate attack hypotheses and try variations at scale, always inside the authorized scope. We also test your own AI products: prompt injection, data leakage and tool abuse.
A specialist on their own tests the hypotheses there is time to test. AI agents walk the whole surface, try variations nobody would write by hand and return candidate flaws. No finding reaches the report without manual reproduction.
Agents enumerate routes, parameters and flows and test each combination. Repetitive work stops limiting how far the test goes.
Agents only reach targets on the authorized list, with rate limits and an agreed window. Destructive actions are off by default.
Every candidate flaw goes to a specialist, who reproduces it, measures the real impact and discards false positives before it reaches you.
If your product uses LLMs or agents, we test what they accept, what they reveal and what they can be led to do, based on the OWASP Top 10 for LLM applications.
We only test systems you own or are authorized to test, under rules of engagement signed before any work starts.
We define targets, test type, window and emergency contacts. Nothing starts without signed rules of engagement.
Surface mapping with AI agents and industry tools. You receive the list of what we found exposed.
Real exploitation attempts inside the scope. Critical findings are reported right away, without waiting for the report.
Every flaw with evidence, reproduction steps, severity and a remediation recommendation. One executive summary and one technical report.
After the fix, we test the findings again and record what was resolved.
An automated scanner report points to hundreds of possibilities. Ours carries only what was reproduced, with the path to fix it.
It is a penetration test in which AI agents do part of the offensive work: reconnaissance, hypothesis generation and attempts at scale. They act only inside the authorized scope, and every finding is reproduced by a specialist before it goes into the report.
Every active test carries some risk, which is why we agree on a window, rate limits and emergency contacts before starting. Destructive and denial-of-service actions are out of scope unless you request and authorize them in writing.
Both. Staging allows more aggressive testing; production shows what is actually exposed. The choice is made during scoping, based on the risk the company accepts.
It depends on the size of the scope: number of applications, APIs, user roles and environments. The timeline is set during the strategy call, before any contract.
The investment depends on the scope and is presented during the strategy call. We do not publish prices on the website.
No test guarantees the absence of flaws. A pentest shows what was exploitable in that scope at that moment, and the retest confirms the fix. Security is continuous, so we recommend testing again after every relevant change.
