Resources
en
Patrowl's AI agents test your entire external attack surface every day, and a certified pentester validates every finding before it reaches you.
Our AI agents link assets, technologies, threats and past tests, and choose the next step.
Our engine runs every repeatable test on every asset, every day.
Certified pentesters prove and validate each finding before it reaches you.
More than 100 clients, including CAC 40 companies
A solution built by pentesters and cybersecurity auditors
Our agents sort through the volume, so your teams only receive findings a pentester has checked and signed.
Platform figures, 2026.
← scroll →
1. Observe
Collect the context around the asset.
Gather the signals around a case.
Patrowl starts from what is actually known about the asset and the threat.
collect asset data, technologies and exposure
pull in vulnerability, CVE and exploit intelligence
bring in test history and previously captured evidence
Result: every decision starts from real, current context, not assumptions.
2. Correlate
Connect signals that belong together.
Connect signals that belong together.
The agent looks for relationships between exposure, vulnerabilities and previous evidence.
link a CVE to the technology in use
connect several findings to the same asset
reuse earlier test evidence when it changes the scenario
Result: related signals become one security question.
3. Decide
Choose the next useful action.
Choose the next useful action.
The agent uses the available evidence to guide what should happen next.
request more context when evidence is incomplete
route a known scenario to automated testing
flag a complex case for expert review
Result: the next step follows the evidence instead of a fixed sequence.
4. Execute or Escalate
Send the action to the right execution path.
Send the action to the right execution path.
Patrowl separates analysis from offensive execution.
run repeatable controls through the testing engine
keep sensitive actions under human decision
move business logic or complex attack paths to a pentester
Result: automation handles repeatable work while experts keep control of high-judgment cases.
The agent uses these layers to guide the next action. It does not replace the testing engine, your security tool stack or human expertise.
| Criterion | Rules-based automation |
On-demand AI pentest |
Patrowl continuous agentic pentest |
|---|---|---|---|
Coverage | All assets, standard checks | The targets you order | Every asset, every day |
Depth of testing | Known checks | Deep on the chosen target | Deep where it changes, continuous everywhere |
Rhythm | On a schedule | Each time you order a test | Every day, plus every release and new threat |
Consistency over time | Repeatable results | A fresh run for each test | Kept until fixed and retested |
Expert validation | Alerts for your team to triage | Depends on the provider | Every finding signed by a pentester |
Time to first results | After setup | After each order | 24 to 48 hours, then every day |
Our research agents study the open-source code your sites rely on. Each discovery is confirmed by a pentester, shared with the maintainer and fixed before publication. Recent CVEs credited to Patrowl:
SQL injection
CubeWP Framework (WordPress)
SQL injection
Frontend Admin (WordPress)
Time-based blind SQL injection
Taskbuilder (WordPress)
SQL injection
Apocalypse Meow (WordPress)
CSRF leading to file deletion
WP Contact Form 7 DB Handler
Open redirect
User Registration & Membership (WordPress)
EXAMPLE OF A ZERO-DAY FOUND BY AI PENETRATION TESTING
Patrowl’s agents hunt unknown flaws in the CubeWP Framework, a WordPress component your assets depend on.
CVE-2026-6453Medium · CVSS 6.5 · CWE-89 · authenticated (Subscriber+). Affects CubeWP Framework ≤ 1.1.30.
The AI brings together asset context, vulnerability intelligence, severity, exploitability, remediation history and previous tests, then decides: automate, gather more context, or escalate.
Replay the relevant control and confirm exploitability.
Complex exploitation and manual validation by an expert.
Reported to the maintainer and fixed upstream. The finding was validated by a pentester before disclosure.
"The race for vulnerability counts is a poor metric: It rewards whoever shows the most, not whoever reduces the risk. The conversation needs to come back to remediation, and that is also the best way to show what triage actually delivers."
Florent Montel – Chief Hacking Officer & Patrowl Co-foundeur
Patrowl combines agentic AI, controlled offensive execution and certified human expertise to keep every test scoped, traceable and safe for production.
Safe testing
AI can recommend the next action, but offensive execution remains bounded by Patrowl’s controlled capabilities, testing rules and production safeguards.
Human oversight
Certified pentesters handle sensitive decisions, complex attack paths and validate the evidence behind each finding.
Security & traceability
Context, decisions, findings and retests remain connected, while sensitive data is protected through French hosting, encryption and enterprise access controls.
Recognition & standards
Patrowl’s testing approach builds on recognized frameworks such as PTES, OWASP, NIST and ANSSI guidance, alongside external industry recognition.
Map and monitor your external attack surface continuously, including assets your inventory has yet to list.
Continuous discoveryDomains, :subdomains, IPs, applications and exposed services.
Shadow IT:Assets linked to your organisation, beyond your current inventory.
Configuration and exposure checks: Services and settings that widen your exposure, flagged as they appear.
Emerging threats:The assets concerned by each new vulnerability, identified within hours.
Automation, agentic AI and certified pentesters test your exposed assets continuously and qualify what can truly be exploited.
Continuous offensive testing: Black-box and grey-box testing of your exposed assets.
Qualification: The real exploitability of each finding, confirmed and signed by a pentester.
Prioritisation: Your teams focus on the risks that matter most.
Remediation and retest:Fix guidance for every finding, and a one-click retest to confirm it.
I need to…
reduce noise and false positive signals between alerts and action
understand why the next test matters
keep AI-assisted decisions traceable
What Patrowl brings: context, evidence and controlled escalation that improve the security posture of security teams.
I need to…
connect threat intelligence with exposed technologies
understand which signal deserves action
keep investigation context when a case escalates
What Patrowl brings: one path from threat signal to validation or expert review.
I need to…
prioritize with more than a severity score
connect asset context with exploit data
route the right issue to remediation
What Patrowl brings: context-aware analysis before the next security action.
I need to…
spend less time on repetitive correlation
receive better context before manual investigation
keep control of sensitive exploitation
What Patrowl brings: AI-assisted preparation without removing human judgment.
I need to…
understand how a change affects an exposed app
send known checks to automation or CI CD pipelines
keep complex application behavior under expert review
What Patrowl brings: adaptive routing between analysis, automation and experts.
I need to…
keep a trace of how findings were analyzed
connect evidence to remediation and retesting
support audit reviews with documented decisions
What Patrowl brings: traceable context and technical evidence throughout the workflow.
Use AI for context, automation for repeatable execution and experts for deeper investigation.
| Title | Asset | Severity | Status | |
|---|---|---|---|---|
| Weak ciphersuite on port 443/tcp | moore.patrowl.io |
Medium | New | |
| Weak protocol on port 8443/tcp | howard-dennis.p… |
Medium | New | |
| Other dangerous service on port 80/tcp | lee-franklin.patr… |
Info | New | |
| Database on port 3306/tcp | elliott.henderson… |
Critical | New | |
| SEG implemented with None | dalton.patrowl.io |
Low | New |
It helps when the next action depends on context. The agent connects asset data, vulnerability information and previous evidence before guiding the next step.
No. Patrowl separates AI guidance from offensive execution. Unlike fully autonomous penetration testing, repeatable tests run through controlled engines, while complex or sensitive actions can move to human experts.
Standard automation works best when the test is already known. AI pentesting adds value when several signals or attack vectors must be connected, or when one result changes what should happen next.
No. Human testers remain important for business logic, complex attack paths, exploit adaptation and sensitive scenarios.
No. Traditional pentesting remains useful for deep human investigation. Agentic AI helps with context, routing and repeatable analysis around that work.
Where an AI-enabled web app exposes relevant attack surfaces, prompt injections can become part of a scoped security assessment. Complex cases should still stay under expert control.
Patrowl keeps the analysis connected to the affected asset, technical evidence, previous test results and the final validation path. The process validates findings before they become actions and keeps validated exploitation separate from AI-only analysis.
Yes. It can add context before automation, expert review, remediation or retesting without replacing the wider security workflow.
