Hadrian
Hadrian Security B.V. · Hadrian: Agentic Attack Surface Management (Atlas) and On-Demand AI Pentesting (Nova)
Hadrian is an Amsterdam-based offensive security platform whose AI agents continuously map an organization's internet-facing assets and confirm which exposures are actually exploitable (Atlas), and run on-demand agentic penetration tests of a single web application (Nova). It suits CISO-led security teams at mid-market and enterprise companies that want validated external findings and pentest evidence for NIS2, DORA, ISO 27001 or SOC 2, with Nova priced from €1,250 per test.
Best for
CISO-led security teams at mid-market and enterprise organizations (Hadrian names 300 to 10,000 employees) with a large, changing internet-facing estate across cloud and on-premises, who want validated external exposures instead of scanner noise and pentest evidence for NIS2, DORA, ISO 27001 or SOC 2
Not ideal for
Teams that need internal network, Active Directory, social engineering or physical testing, financial entities looking for their DORA TLPT, organizations that require self-hosted tooling, developers who want an open-source pentesting agent to run in CI, and small companies without a dedicated security team
Who it's for
CISOs, security operations, vulnerability management, offensive security and GRC teams at mid-market and enterprise organizations
Capabilities
- Atlas external attack surface discovery from no starting scope: domains, subdomains, DNS records, IPs, certificates, cloud services and apps, including assets nobody remembers owning
- Continuous, event-driven testing: passive discovery every hour, and new tests whenever a host, port or software version changes
- Exposure validation by AI agents before alerting: each reported risk carries a proof of concept, reproduction steps and remediation guidance
- Emerging threat response: new CVEs are matched to assets running affected versions and tested against them
- Prioritization by business context, asset criticality, CISA KEV data and threat intelligence rather than CVSS or EPSS alone
- Nova on-demand agentic pentests of one application: reconnaissance, exploitation, privilege escalation, authenticated and IDOR testing, cross-tenant checks, business logic and attack chains
- Live view of Nova's reasoning, the tests it ran and the evidence collected, with every request and response attached to findings
- Compliance-ready Nova PDF report mapped to SOC 2, ISO 27001 and NIS2, with results Hadrian describes as human-reviewed
- Automatic retesting when a risk is marked resolved; the risk closes only if the exposure can no longer be exploited, and each Nova pentest includes unlimited retests for 12 months
- Safety controls described by Hadrian: customer-set rate limits, a shared request budget across agents, no destructive HTTP methods such as DELETE, credentials hidden from the model, and human safety checks before higher-risk testing
- Atlas add-ons for mergers and acquisitions assessment and infostealer credential leak detection with dark web monitoring
- Integrations with Jira, ServiceNow, Azure DevOps, Linear, GitHub, Slack, Microsoft Teams, PagerDuty, Splunk, Microsoft Sentinel, Google Security Operations, Cortex XSOAR, Tines, CrowdStrike Falcon, Microsoft Defender and SentinelOne, plus webhooks and the Hadrian API
- Cloud connectors for AWS, Microsoft Azure, Google Cloud and Cloudflare that bring new public-facing resources into scope as they appear
Limitations
- External attack surface only: Hadrian's FAQ says internal networks, social engineering and physical access still need a human red team
- Each Nova pentest covers one target: marketing pages mention web apps, APIs and cloud infrastructure, but the Nova terms define a target as one external web application identified by a single URL, so several apps mean several tests or a bundle
- Nova pentest entitlements are used up when a test starts and are not refunded after that unless Hadrian itself aborts the test; unused entitlements expire at the end of each contract year and do not roll over
- Out-of-scope guidance is free text that Nova treats as guidance only, and the terms say Hadrian does not warrant strict compliance with it
- You must own each target or hold every necessary authorization, including from hosting providers and other tenants reached by cross-tenant tests, and the terms make you indemnify Hadrian for unauthorized targets
- Atlas has no published price; it is quoted on total asset count, and its add-ons cost extra
- Hadrian's terms say it does not warrant that Nova will find every vulnerability, and that AI-generated output may contain errors
- Hadrian says it does not replace a DORA threat-led penetration test (TLPT) or certify compliance; compliance-framework references describe the controls Nova helps evidence
- No public product or API documentation is published beyond the website, FAQs and terms; the Nova terms refer to documentation provided to customers
- No self-hosted or on-premises deployment is documented; it runs as a cloud service
- Performance figures such as noise reduction, remediation speed and ROI are Hadrian's own claims
Use cases
- Building and maintaining an inventory of internet-facing assets, including shadow IT and subsidiaries, for a CISO-led security team
- Checking within hours whether a newly disclosed critical CVE affects any exposed asset, with proof for the ones that are exploitable
- Running an agentic pentest of a customer-facing web app on staging before a launch instead of waiting for a pentest firm's calendar
- Testing whether one customer's account can reach another customer's data in a multi-tenant SaaS app
- Producing pentest evidence for a SOC 2, ISO 27001 or NIS2 audit, and retesting fixes before the auditor arrives
- Assessing the external attack surface of an acquired company during due diligence with the M&A add-on
Our take
Hadrian is aimed at the security team that already owns a large, untidy internet footprint and is drowning in scanner output. Atlas's value is the inventory plus the filter: it keeps finding assets and only alerts on what its agents could exploit. Nova then gives that team a priced, on-demand pentest of a specific app, with evidence and retests built in. The combination is the distinct part; most agentic pentesting tools in this list either test one app you point them at or are built for developers in CI. The limits are clear from Hadrian's own pages: external surface only, Atlas pricing by quote, one URL per Nova test with yearly-expiring entitlements, and little public documentation. Treat the vendor's noise and ROI figures as something to confirm in a demo against your own attack surface.
Who should use it
CISOs and security operations, vulnerability management and GRC teams at mid-market and enterprise organizations with many internet-facing assets, subsidiaries or acquisitions, who want a continuously validated external inventory and on-demand pentests with audit-ready evidence, especially in the EU.
Who should skip it
Teams whose main risk is internal (Active Directory, lateral movement inside the network) or human (phishing, physical access), financial entities that need a TIBER-EU or DORA TLPT, organizations that must self-host security tooling, and developers who want an open-source pentesting agent they can run themselves.
Strengths
- Continuous external discovery and on-demand pentesting in one platform with shared context
- Findings are validated before alerting, with proof of concept, reproduction steps and remediation guidance
- Published starting price for Nova tests, with unlimited retests for 12 months
- Detailed public description of Nova's safety controls and human review
- EU data storage and EU model inference stated in the Nova terms
- Broad ticketing, SIEM/SOAR, endpoint and cloud integrations
Weaknesses
- External attack surface only; no internal network or social engineering testing
- Atlas pricing is by quote only
- One Nova test covers one web application URL, and unused entitlements expire yearly
- Little public product or API documentation
- Performance and ROI figures are vendor-reported
Hadrian pricing
Nova
From €1,250 per test (excl. VAT)
- On-demand agentic pentest of one target (a web application at a single URL)
- Proof of exploitability, reproduction steps and remediation guidance for each finding
- Compliance-ready PDF report mapped to SOC 2, ISO 27001 and NIS2
- Unlimited retests within the first 12 months
- Bundles available for multiple tests
Atlas
Custom (based on total asset count)
- Continuous external attack surface discovery and exposure validation
- Hourly passive discovery and event-driven testing
- Add-ons at extra cost: mergers and acquisitions assessment, infostealer credential leak detection
Note: Nova is bought through an order form for a subscription period; each test consumes one pentest entitlement when it starts, consumed entitlements are not refunded unless Hadrian aborts the test, unused ones expire at the end of the contract year, subscription periods auto-renew for 12 months unless either side gives 30 days' notice (unless the order form says otherwise), and the terms allow an annual fee increase in line with EU consumer price inflation (HICP). Hadrian also offers a free external exposure report, which is not a pentest.
Technical specs
Available models
Where Hadrian excels
Getting control of a sprawling external footprint
Atlas starts without a scope list, finds forgotten hosts, cloud resources and subsidiary assets, and only alerts on exposures its agents could prove, which turns a scanner backlog into a short, evidenced queue.
Same-day answer to a critical CVE
New CVEs are matched to exposed assets running the affected version and tested, so the team knows which systems are really exploitable without waiting for a manual check.
Pentest evidence on the release schedule
A Nova test can be launched against a staging app before a release or audit, usually finishes within one to two days, and includes retests so fixes can be confirmed before the report goes to the auditor.
Hadrian vs. competitors
Hadrian vs. Fleuret
Fleuret is a Paris-based agentic pentesting service priced at a flat €4,000 per web app, covering the app, its API and the external infrastructure behind it, with a free-to-start scanner plan for code and domains. Hadrian's Nova is the closer match and starts at €1,250 per test, while Hadrian adds Atlas for continuous discovery and validation across the whole external attack surface. Both say pentest data and model inference stay in the EU.
Hadrian vs. Strix
Strix is an Apache-2.0 open-source pentesting agent that developers run themselves in the CLI or CI with their own model, plus a paid cloud platform for scheduled pentests and PR reviews. Hadrian is a closed, managed platform for security teams: continuous external attack surface management with Atlas and human-reviewed Nova pentests, sold through sales with ticketing and SIEM integrations.
Frequently asked questions
What does Hadrian do?
Hadrian is an agentic offensive security platform with two products. Atlas continuously discovers an organization's internet-facing assets and uses AI agents to confirm which exposures are exploitable before alerting. Nova runs on-demand agentic penetration tests of a single application, with every finding backed by the requests and responses that prove it.
What is the difference between Hadrian Atlas and Nova?
Atlas covers breadth: it maps and tests your whole external attack surface continuously, with hourly passive discovery and new tests on every change. Nova covers depth: you point it at one application and its agents pentest it in detail, including authenticated, IDOR, cross-tenant and business logic testing. They share context, so Atlas can surface what deserves a deeper Nova test.
How much does Hadrian cost?
Nova starts at €1,250 per test excluding VAT, with bundles for multiple tests and custom quotes based on scope and frequency. Atlas is priced on your total asset count by quote, and its mergers and acquisitions and infostealer credential leak add-ons cost extra. Each Nova pentest includes unlimited retests within the first 12 months at no extra cost.
How long does a Nova pentest take?
Hadrian says most Nova tests finish within 24 to 48 hours, depending on scope. You can follow results in the platform as they come in and receive a compliance-ready PDF report at the end.
Are Hadrian's findings reviewed by humans?
Hadrian describes Nova results as human-reviewed. Its engineering write-up says findings pass automated verification routines and that novel edge cases are routed to human security practitioners, and its FAQ says humans stay in the loop for scoping and run safety checks before higher-risk testing.
Is it safe to run Hadrian against production systems?
Hadrian says it uses safe exploitation, that you approve the scope before a test starts, and that its hackers run a safety check before higher-risk testing. Nova lets you set rate limits, removes destructive HTTP methods such as DELETE from its agents' toolkits, and hides test credentials from the model. You can also run Nova against a staging environment first.
Does Hadrian satisfy DORA or NIS2 requirements?
Hadrian supports the external attack surface parts of these frameworks, and Nova reports map to SOC 2, ISO 27001 and NIS2. Hadrian states that it does not certify DORA compliance or replace a DORA threat-led penetration test (TLPT), and its terms say framework references describe the controls Nova helps you evidence.
Where does Hadrian process data, and which AI models does it use?
Nova's terms say all customer data is stored and processed in the EU and that foundation-model inference happens in the EU, and that customer data is not used to train third-party models. Hadrian's sub-processor list names Microsoft Azure, through which it accesses OpenAI models hosted in the EU. For the wider platform, the list shows product data stored in the EU, with an opt-in for customers who want scanning from other regions.
Integrations & fit
Alternatives to consider
Fleuret AI (FLEURET AI SAS)
European companies of roughly 50 to 1,000 people with a real web and API attack surface (SaaS, fintech, healthtech) and a NIS2, DORA, SOC 2 or ISO 27001 deadline, especially those that want pentest data and model inference kept on EU infrastructure
Freemium
Strix (OmniSecure, Inc.)
Engineering and AppSec teams that want open-source, AI-driven pentesting of their own apps and APIs in the CLI or CI, with an option to move to a managed platform
FreemiumAbout Hadrian
Hadrian, founded in Amsterdam in 2021 by former teenage ethical hackers Rogier Fischer and Olivier Beg together with Maurice Clin, sells two products on one platform that it positions as adversarial exposure validation for continuous threat exposure management (CTEM) programmes. Atlas covers breadth: it starts from no scope, discovers the domains, DNS records, IPs, certificates, cloud services and applications a company exposes online (including shadow IT), runs passive discovery every hour, and triggers new tests whenever something changes, such as a new host, port or software version. When a new CVE is published, Atlas matches it to assets running the affected version and tests them. Its agents try to exploit each potential exposure before raising it, so the team receives a short list of verified risks with proof of concept, reproduction steps, business-context prioritization beyond CVSS or EPSS, and remediation guidance, instead of a scanner backlog. Nova covers depth: you point it at one application, and a fleet of agents signs up like a real user, maps the API, and works through reconnaissance, exploitation, privilege escalation, authorization (IDOR) and cross-tenant testing and business logic, chaining small flaws into demonstrated impact. Hadrian says most tests finish within 24 to 48 hours, that you can follow the AI's reasoning, tests and evidence live, and that results are human-reviewed; the test ends with a compliance-ready PDF report mapped to SOC 2, ISO 27001 and NIS2, and each Nova pentest includes unlimited retests for 12 months. Hadrian's engineering write-up describes Nova as a layered system rather than one free-roaming agent: deterministic open-source tools do reconnaissance, a coordinator spawns narrowly scoped attack agents, a supervisor agent shifts budget toward promising attack paths, agents share a rate budget, destructive HTTP methods such as DELETE are removed from their toolkits, credentials are hidden from the model behind an abstraction layer, and novel edge cases go to human practitioners for review. Both products run from the cloud with nothing to install, feed findings to ticketing, chat, SIEM/SOAR and endpoint tools, and connect to AWS, Azure, Google Cloud and Cloudflare for cloud asset discovery. Nova's terms state that customer data is stored and processed in the EU and that foundation-model inference happens in the EU. The tradeoffs: Hadrian only tests the external, internet-facing attack surface, so internal networks, social engineering and physical red teaming still need people; Atlas is priced on total asset count by quote; each Nova test covers one target defined by a single URL, unused entitlements expire at the end of each contract year, and the customer must warrant authorization over every target; and no public product or API documentation is published beyond the website, FAQs and terms.
Updates from Hadrian
Hadrian announced a $40M round co-led by Forgepoint Capital International and SmartFin, with HV Capital, Motive Partners, Picus Capital and Oetker Ventures participating, bringing its total funding to $65M. It says the money will fund expansion across EMEA and the U.S. and more investment in its engineering and research teams behind Atlas and Nova.
Hadrian launched Nova, an on-demand agentic pentesting product that adds deep testing of specific applications to its external exposure management platform. It is priced per test rather than as a recurring retainer, and Hadrian describes the results as human-reviewed.
Are you the founder? Claim this listing →