Skip to main content

AI penetration testing. Privately, on your hardware.

Excalibur's Agentic Pentesting module attacks your web apps, cloud, APIs, internal network and external perimeter continuously. Then it waits for a person to approve every critical step. Your data never leaves your environment.

AGENT · session a7f3e

Live
Hypotheses
43generated
Exploits
7verified
Boundaries
2held
Approval
1pending
  1. ▸ reconnaissance
  2. · 47 endpoints · 3 auth schemes
  3. · 12 hypotheses prioritized
  4. · probe /api/users?id= → SQLi
  5. ✓ VERIFIED time-based blind SQLi
  6. ▸ lateral movement
  7. ✓ extracted aws_access_key_id
  8. ✓ IMDSv1 exposed → IAM role
  9. ▸ impact assessment
  10. ⊘ BLOCKED · pentest-scope-v4.2 §3.1
  11. ▸ chain assembled · CVSS 9.8
  12. · SQLi → creds → SSRF → IAM
  13. ⧗ requesting approval @sarah.k

Product interface shown with illustrative sample data.

Most security teams describe the same gap. “We pentest once a year, if that. Between tests, nobody knows what an attacker could actually get into.”

Every surface a pentest team would test.

One agentic capability for web, cloud, API, internal and external assessment. It produces verified, contextualized vulnerabilities, not a list of maybes.

▸ reconnaissance

· 47 endpoints

· 3 auth schemes

· 12 hypotheses prioritized

External pentesting

Probes your internet-facing perimeter the way an outside attacker would, and keeps probing as new services appear.

Recon

Initial access

Privilege escalation

Lateral movement

Internal pentesting

Runs Atomic Red Team tests and MITRE ATT&CK simulations inside your network, within the scope you control.

SQLi

→ creds

→ SSRF

→ IAM · CVSS 9.8

Cloud and API testing

Tests endpoints, auth schemes and cloud permissions, then chains what it finds into the path a real attacker would take.

Machine speed. Human approval.

The agent moves fast, but it never acts alone on what matters. When it finds something critical, it stops and asks a named person before anything is submitted.

  • You set the scope. The agent holds every boundary in it.

  • Critical findings wait for approve, reject or request info.

  • Every step is logged, so your team can check how a finding was made.

Approved@sarah.k

Probe: SQLi on /api/users

Boundary heldscope-v4.2 §3.1

Attempt to read production PII

Awaiting approvalcritical · CVSS 9.8

Submit finding: SQLi → SSRF → IAM

Built for networks that can't phone home.

Excalibur ships on NVIDIA GPU hardware into your datacenter and runs CyberAGI's own models. Your targets, findings and exploit chains stay inside your four walls.

No phone-home. No telemetry. No license check.

CVE corpus, nuclei templates and TTP playbooks pre-loaded at delivery.

You control inbound sync from NVD and EPSS. Nothing flows out.

CapabilityExcaliburTypical AI security
Data sovereigntyOn-prem, air-gappedCloud or hybrid
ModelsCyberAGI's own models, no API dependencyOften a frontier model behind an API
ScopeEnd-to-end lifecycleMostly validation
Human controlBuilt-in HITL checkpointsLimited
Model evolutionPrivate continuous fine-tuningNone

A pentest that feeds the rest of your program.

“Do the pentest, the SOC alert and the threat model share one context, or three databases behind one login?” In Excalibur, all nine modules share one Security Knowledge Graph. A verified exploit chain moves straight into the stages that act on it.

Correlate

Exposure Fabric

Verified findings join what Qualys, Tenable, Wiz and your other scanners report, deduplicated into one view.

Remediate

Reporting

Each finding becomes part of a living, editable report your team can share and act on.

Govern

Compliance

Findings map to controls, so pentest evidence is ready before the audit asks for it.

Learn

Self-Evolving AI

Models fine-tune on your private data, on your hardware, and get sharper every cycle.

Pentesting alone costs more than the whole platform.

CyberAGI's pricing page puts standalone pentesting and breach simulation at $175,000 a year. Excalibur includes Agentic Pentesting and eight other modules for one fixed cost.

Pentesting / BAS today

$175,000/yr

Typically Cobalt, NetSPI, Pentera or AttackIQ

Excalibur

$80,000/yr flat

Hardware included. All nine modules. No priced seats.

“I was initially skeptical about AI's impact on the cybersecurity industry, until I started working with Excalibur.”

Excalibur’s modules and technology have significantly improved my ability to manage multiple security frameworks and architectures. The Posture model delivers actionable insights that enhance the effectiveness of Tier 1 analysts while providing an excellent learning path, reducing the time I need to spend on foundational training. It also does an outstanding job of modeling potential threats and identifying security concerns.

The recently introduced Govern model is ahead of the curve when it comes to reducing the time required for compliance management and audit preparation. The ability to answer framework-specific questions simply by uploading policy documents is a tremendous time saver. Our previous platform, Apptega, which remains a solid solution in terms of cost and capabilities, could not extract answers directly from documentation, making this feature a true game changer.

I firmly believe AI should be used to make security professionals more effective and efficient, freeing them to focus on meaningful security work rather than replacing jobs. Excalibur fully embraces that philosophy and continues to deliver rapid innovation and improvement.

I’m excited to see where this product goes next.

Loial (A Logicalis Company)

Cody Jackson

Principal Engineer, Loial (A Logicalis Company)

Questions buyers ask about AI pentesting.

AI penetration testing uses AI agents to do the work of a pentest: map targets, form attack hypotheses, try exploits and verify which ones actually work. Excalibur's Agentic Pentesting module does this continuously across web, cloud, API, internal and external targets, instead of once a year.

Nine Teams. One closed loop.

Discover. Model. Validate. Correlate. Posture. Remediate. Govern. Predict. Learn. Each stage sharpens the next. See what happens when they close the loop.

Already know what you need? Configure Excalibur