NodeZero, from Horizon3.ai is an autonomous penetration testing platform. Your team runs it against your own internal network, Active Directory, AWS and Azure accounts, Kubernetes clusters and external perimeter, and since late July 2026 against web applications, and it reports the attack paths it proved. It fits security teams that want to run unlimited tests themselves across the year. If you need an outside team to run the pentest and hand you a SOC 2 and ISO 27001-ready report within 48 hours, CodeAnt AI is the better fit.
CodeAnt AI has run 12,000 pentests and is trusted by more than 1,000 companies, including Fortune 500 companies like NRG, cybersecurity leaders like Commvault, and fast-growing companies like 11x and Sanas.
Every Horizon3.ai and competitor fact below was checked on the vendor's own page or documentation on that date.
What Does NodeZero Include in 2026?
Horizon3.ai sells NodeZero in four packages priced per asset, plus a web app product priced per application. From its pricing page, checked November 12, 2026:
Every package: internal, external, cloud (AWS and Azure) and Kubernetes pentests, Active Directory password audits, insider threat and phishing impact testing, and Fix Actions with 1-click Verify to retest a fix.
Flex, the entry package, also lists external asset discovery, comprehensive reports, the NodeZero MCP Server, the Vulnerability Management Hub and Endpoint Security Effectiveness.
Core adds scheduling and Threat Informed Perspectives.
Pro adds Tripwires and Rapid Response.
Elite adds Risk-Based Exposure Management, Vulnerability Prioritization, High-Value Targeting, Advanced Data Pilfering, Threat Actor Intelligence and NodeZero Insights.
WebApp: sold as an add-on or on its own, per application.
Horizon3.ai gates packages on how often you test and on detection and intelligence features, not on test types. Nothing on this list reads source code.
What Features Does NodeZero Actually Include?
NodeZero runs one kind of test across many surfaces. Each test launches a safe, controlled attack and shows you the path an intruder would take. A single portal pulls the findings into deduplicated, impact-scored attack paths with 1-click verification of fixes.
Here is each module the way you meet it after your first pentest, starting with the network tests that anchor the platform.

Internal network pentest
Internal Pentesting is the flagship. NodeZero autonomously discovers and exploits weaknesses inside your network through lateral movement, credential attacks, data mining, security-control bypass, and vulnerability exploitation.
The engine goes far beyond CVEs into misconfigurations and chained weaknesses, then post-exploits with credential dumping and RAT implantation. Every result arrives as proof, so you see the exact path to domain compromise instead of a theoretical score.
External network pentest
External Pentesting shows your organization through an attacker’s eyes with a “Discover, Authorize, Pentest, Repeat” workflow. Passive asset discovery leverages DNS and OSINT to enumerate public-facing systems, and an authorization step controls which of them get tested.
One staleness note from the release notes: the 2026.06 rebuild replaced Asset Groups with reusable Scopes, and the marketing copy still lags that change. Trust the docs over the product page here.
Cloud pentest for AWS and Azure
Cloud Pentesting takes a gray-box approach into AWS and Azure. For AWS, NodeZero enumerates resources to find an opening, then uses privilege escalation and lateral movement. For Azure, it blends native attacks with harvested data to pivot in and out of hybrid environments.
Horizon3.ai claims autonomous escalation to Microsoft Entra ID Global Admin without relying on CVEs. The docs expose an AWS Pentest and an Azure Entra ID test as distinct test types rather than one generic cloud button.
Kubernetes pentest
Kubernetes Pentesting runs against live clusters and hunts RBAC misconfigurations, container escapes, and secret exposures. It deploys inside clusters using Kubernetes Operators and infrastructure-as-code, and chains findings with the underlying infrastructure.
Managed platforms are covered, including AWS EKS, Google GKE, and Azure AKS. Testing happens at runtime against the cluster as it actually runs.
Active Directory testing
Active Directory is where NodeZero earns its reputation, on the premise that “attackers don’t hack in, they log in.” AD Password Audit cracks passwords that are exposed to spraying, stuffing, and cracking using public breach data and OSINT tied to your company, then ranks the riskiest accounts and their blast radius.
The identity story has a detection half too. AD Tripwires plant decoy accounts that look exploitable but stay uncrackable, and they fire on credential abuse like Kerberos ticket harvesting and AS-REP roasting.
Rapid response for emerging threats
Rapid Response answers a specific fear. Attackers weaponize exploits in hours, while patch cycles still run weeks. The Horizon3 attack team builds production-safe exploits for fresh CVEs and KEVs, often within hours of disclosure, so you get proof of exploitability rather than a version check.
As Horizon3.ai puts it, “exploitability is confirmed or ruled out with evidence, not vendor advisories or CVSS scores from vulnerability scanners that just check versions.” Recent 2026 coverage spans Esri ArcGIS Server, Adobe ColdFusion, SimpleHelp, and Apache ActiveMQ. Rapid Response starts at the Pro tier.

Tripwires and deception
NodeZero Tripwires turn pentest findings into defense by dropping expert-designed decoys as the engine discovers critical risk. The decoys and monitoring files behave as honeytokens, staying silent until an attacker touches them, then firing an immediate alert.
Placement is automatic on your crown-jewel assets, and one-click activation works inside internal, external, AWS, phishing, insider-threat, or Rapid Response tests. Alerts flow into your SIEM and SOAR through Splunk and Microsoft Sentinel webhooks.
Phishing impact testing
Phishing Impact Testing answers what actually happens after an employee gets phished. Rather than measuring click rates, NodeZero captures the phished credential and then tests what an attacker could do with it through lateral movement, privilege escalation, and data access.
Horizon3.ai frames it as “Beyond Simulation: Proven Impacts,” and positions it to supplement your existing simulated-phishing tool. The output is the reachable blast radius, not a training score.
Endpoint security effectiveness
Endpoint Security Effectiveness validates whether your EDR earns its budget. NodeZero deploys a test remote access tool, simulates attacker behavior in the live environment, and reports whether the EDR blocked, alerted, or missed the activity.
CEO Snehal Antani is blunt about why, noting that “our research shows that credential-based attacks can bypass EDRs in minutes, often undetected.” Every customer gets the feature, including the Flex tier, though Horizon3.ai names no specific EDR vendors.
NodeZero Insights
NodeZero Insights is the executive layer, giving organization-wide trend data so a leader can prioritize and prove impact. It aggregates mean-time-to-remediation, open weaknesses and attack paths over time, weakness age, and pentest-series trends.
Insights is Elite-tier only, which places board-level reporting at the top of the four-tier ladder. Everything below it can run the tests but cannot roll the results into a strategic report.
Web application pentest (NodeZero WebApp)
Horizon3.ai added web application pentesting to NodeZero in late July 2026. The NodeZero WebApp page says it can "Test authenticated, role-based workflows with credential and MFA support", discovers single-page apps and REST, SOAP and GraphQL APIs "using headless browser crawling", and checks for business logic flaws, broken access control, IDOR and BOLA. It offers "graduated testing modes" for production, staging and development, and each finding comes with "replayable proof, screenshots, and request/response evidence".
The point Horizon3.ai stresses is chaining: linking an application flaw to credential theft, lateral movement, cloud access or data exposure elsewhere in your environment. WebApp is priced per application: WebApp Flex at $10,000 per app, per test, or WebApp Continuous at a custom price with unlimited tests of each licensed app.
The AI architecture and MCP server
NodeZero’s AI rests on graph-based reasoning, deterministic logic, and scoped GenAI across multiple cooperating agents such as the Exploit Suggester, High-Value Targeting, and Advanced Data Pilfering. Horizon3.ai is explicit that “NodeZero doesn’t train foundation models,” and instead builds structured prompts from live data and runs inference against models like Claude, LLaMA, or Mistral.
The NodeZero MCP Server extends that reasoning to your own agents. It acts as a natural-language bridge between your LLM and the NodeZero API, letting you launch pentests and query asset and vulnerability data from an assistant, in either a Horizon3-hosted mode with OAuth 2.1 or a self-hosted single-user mode.
What does each NodeZero test need from you?
NodeZero installs no agents on your endpoints, but most tests need something from you. From Horizon3.ai's documentation, checked November, 2026:
Test | Where it runs | What you provide |
|---|---|---|
Internal pentest | A Docker host inside your private network | The host, placed in the segment where the test should start |
Entra ID hybrid pentest | A Docker host inside your private network | The host |
External pentest | Horizon3.ai's cloud | A scope, with the assets you authorize |
Kubernetes pentest | Inside the cluster | Operator and ClusterRole permissions |
AWS pentest | Horizon3.ai, through your account | A privileged role in your AWS account |
Azure cloud pentest | Your Azure subscription | A managed identity |
Phishing impact test | Part of an internal pentest | Phished credentials to inject |
The NodeZero host runs on Ubuntu 20.04 or later or RHEL 9 or later, with Docker 20.10 or later or Podman 4.0 or later. The minimum is 2 cores, 8 GB of RAM and 40 GB of free disk; Horizon3.ai recommends 4 cores and 16 GB of RAM. Its documentation says not to run EDR on the host, because EDR can stop NodeZero, and the product page offers the host as a free Docker host or virtual appliance (OVA).
Who Runs NodeZero, and Does it Count as Your Annual Pentest?
NodeZero is a platform your team operates. You set up the host, create scopes, launch or schedule tests and act on the results. That is the appeal: on the annual packages you can test as often as you like. It also raises the question auditors and enterprise customers ask first: who did the testing?
PCI DSS v4.0.1 is the clearest example. Requirements 11.4.2 and 11.4.3 call for internal and external penetration tests at least once every 12 months and after any significant change, performed "by a qualified internal resource or qualified external third-party", with "organizational independence of the tester" (the tester does not have to be a QSA or ASV). Running NodeZero yourself can support that, but the person running it has to be qualified and independent of the systems under test.
Horizon3.ai's own answer for PCI is a service, not the self-run platform. Its PCI page says "Horizon3 Offensive Security Certified Professionals (OSCPs) deliver penetration testing services" for Requirement 11.4, with a penetration test report and a Fix Action report that can be shared with your auditor, plus NodeZero access to verify fixes.
For SOC 2 and enterprise security reviews, ask your auditor or customer before you plan to use results from a tool you run yourself as your yearly third-party pentest. If they want an outside firm's report, budget for that separately.
What’s Good About NodeZero’s Feature Set?
Proof over theory. NodeZero shows the path it took, not a list of CVEs; Horizon3.ai's line is "NodeZero isn't a scanner." Fix Actions with 1-click Verify let you rerun the exact attack after a fix, on every package.
Volume. Core, Pro and Elite count each asset once a year and allow unlimited tests, so a team can test internal, Active Directory, cloud and Kubernetes after every change without a new purchase order.
Reach across your environment. One platform covers internal networks, identity, cloud, Kubernetes, the external perimeter and now web apps, and chains findings across them. That is the part most web-only AI pentest tools do not do.
Production use. Horizon3.ai says NodeZero "has achieved zero downtime across all production tests", and its homepage lists 7,012 customers (checked November, 2026).
Where Does NodeZero Stop?
You operate it. Someone on your team has to set up hosts, scopes and schedules, triage results and drive fixes. Teams without that time get less from unlimited testing.
Internal tests need a host in your network. Internal and Entra ID hybrid tests run from a Docker host you place and maintain, with EDR kept off it.
It tests what you scope. External tests run against the assets you authorize. Subdomains on your domain that a vendor hosts for you are only covered if you put them in scope, and some vendor contracts limit who may test them.
No source code. NodeZero tests running systems; it does not read your repositories, so it cannot see a flaw that only shows up in code or configuration you have not deployed.
Web app tests are priced per app, per test. WebApp Flex is $10,000 per app for each test; unlimited testing needs WebApp Continuous at a custom price.
No report deadline. Horizon3.ai's pages say you can start a test in minutes, but none of the pages we checked commits to when a report is ready.
How Do NodeZero’s Features Compare With Other Ways to Get a Pentest?
NodeZero owns the network, infrastructure, identity, and threat-detection layer, and CodeAnt AI owns the code and application layer with a free way to begin.
Where NodeZero reads none of your source, CodeAnt AI’s pentest is code-aware and reads it in black, white, and grey box modes to surface authorization and logic flaws a network attack cannot see. CodeAnt also folds in SAST, SCA, secrets, and IaC, plus AI code review and DORA delivery metrics, none of which NodeZero offers.



Here is where each platform’s coverage falls, feature area by feature area.
Provider | Model | Time to report | Pricing model | Report and compliance mapping | Best for | Main tradeoff |
|---|---|---|---|---|---|---|
CodeAnt AI | Pentest run for you: AI agents attack your systems and a researcher checks every critical finding. Black, grey and white box across web apps, APIs, cloud, Kubernetes, Active Directory, internal and external networks | Within 48 hours | Priced per pentest on the attack surface you have, with generous pricing for fast-growing companies | SOC 2 and ISO 27001-ready report; unlimited retest for 12 months | Teams that need an outside pentest and an auditor-ready report fast | No public list price; you get a quote for your scope |
Horizon3.ai NodeZero | Autonomous platform your team runs: internal, external, AWS and Azure, Kubernetes, Active Directory, phishing impact; web apps as a separate product | No report deadline stated; tests start in minutes | Per asset: Flex $15 per test; Core $50, Pro $65, Elite $85 per year with unlimited tests; WebApp Flex $10,000 per app per test | Comprehensive reports on every package; separate PCI 11.4 pentest service delivered by Horizon3.ai's OSCP-certified staff | Teams with staff to run frequent internal, identity and cloud tests | You operate it; internal tests need a host in your network; no source code |
Pentera | Automated security validation platform (internal, external, cloud and hybrid, web apps and APIs) with optional expert-led testing | Not stated | Not public; demo and quote | "Controlled execution with audit proof"; no framework mapping stated on the homepage | Enterprises validating security controls across a large estate | No public price or report deadline |
XBOW | Autonomous AI pentest of web apps and their APIs; "No human drives the attack", and findings are reviewed before delivery | First validated findings in hours | Usage-based, quote only; sold on AWS, Google, Oracle and Microsoft marketplaces | Homepage lists SOC 2, ISO 27001, ISO 42001, PCI DSS and NIS 2; free unlimited retests | Many web apps tested continuously | Web apps and APIs only; no public price |
Cobalt | PTaaS with vetted human testers on credits (1 credit equals 8 hours of testing), plus an autonomous pentest directed by a Cobalt pentester | Autonomous: findings within 24 hours; human tests start in 1 to 3 business days | Autonomous pentest $3,500 per test, a limited offer for tests completed before December 31, 2026; credit price not published | Tiers positioned for compliance needs; unlimited retesting during the contract | Teams that want human testers on call through the year | The $3,500 price is time-limited; credits bought upfront |
For a longer list of options, see our full head-to-head CodeAnt AI vs NodeZero breakdown.
NodeZero or CodeAnt AI: Which Fits?
Choose NodeZero when your team wants to own the testing: run unlimited internal, Active Directory, cloud and Kubernetes tests after every change, and has people to read the results and drive fixes.
Choose CodeAnt AI when you want the pentest done for you: an outside team runs it, a researcher checks every critical finding, and you get a SOC 2 and ISO 27001-ready report within 48 hours. CodeAnt also runs white box tests with code access, which NodeZero does not offer.
Some teams run both: NodeZero for frequent internal testing they operate themselves, and an outside pentest for the report their auditor or customers ask for.
What Do Buyers Ask Before They Choose NodeZero?
These come up on our own sales calls, paraphrased:
"Would it have found what an outside tester found?" Teams that run NodeZero inside the network worry most about the outside: portals and subdomains on their domain that a vendor hosts, which were never in an authorized scope. Before you rely on any tool for external coverage, list every host on your domains, including vendor-hosted ones, and check which are in scope.
"Can we compare two tests fairly?" Run both against the same scope in the same week, then compare exploited findings with their evidence (the request, the response and the steps to reproduce), not the number of findings.
"Who signs the report?" Auditors and enterprise customers often want an outside firm's name on the annual pentest. If that is your situation, a platform you run yourself does not replace it.
"Is it safe on production?" Ask any vendor how it avoids changing real data, how you stop a test mid-run, and whether every request is logged.
How Does CodeAnt AI Run a Pentest?
CodeAnt prices each pentest on the attack surface you have, and gives very generous pricing to fast-growing companies. You get a quote for your scope before the test.
What we test: black box, grey box and white box pentests for web apps, APIs, cloud, Kubernetes, Active Directory, and internal and external networks.
How findings are proven: AI agents attack your systems the way an adversary would, findings are reproduced in a sandbox, and a researcher checks every critical finding before it reaches you. Every request, response and decision point is logged in sequence.
What you get: a SOC 2 and ISO 27001-ready pentest report within 48 hours, and unlimited retest for 12 months. We run the same attack again after every fix, so you know the fix held.
Proof behind the team: SOC 2 Type II, Gartner Cool Vendor 2026 and CERT-In empanelment. As a VulnCheck CNA partner we published CVE-2026-29000 (pac4j-jwt, CVSS 10.0) and CVE-2026-28292 (simple-git, CVSS 9.8), both missed by Snyk, Semgrep, SonarQube and Checkmarx.
CodeAnt AI has run 12,000 pentests and is trusted by more than 1,000 companies, including Fortune 500 companies like NRG, cybersecurity leaders like Commvault, and fast-growing companies like 11x and Sanas.
Get a quote for your scope at codeant.ai/pentesting.
Where This Leaves You
NodeZero is one of the broadest autonomous pentest platforms you can run yourself: internal networks, identity, cloud, Kubernetes, the external perimeter and now web apps, with unlimited tests on its annual packages. Its limits follow from that model. Your team operates it, internal tests need a host, it tests only what you scope, and it does not read code. If what you need this quarter is an outside pentest and a report your auditor or customer will accept, start there, and use a platform like NodeZero for the testing in between.
The other tools are complementary, and the practical move is to pick by layer. For pricing detail, read the NodeZero pricing breakdown, and for other options, see the NodeZero alternatives roundup. If the code and app layer is what you need to cover, CodeAnt AI is the one you can evaluate for free today.







