Quant Sentinel

Automated penetration testing, powered by AI

Specialist AI agents orchestrated across adaptive pipelines. White-box, grey-box, or black-box testing for web applications, APIs, and network infrastructure — with exploitation proof for every finding.

Sample output

See what Sentinel delivers

Findings summary with proof statuses, detailed exploitation evidence, a deterministic findings roster, and per-CWE remediation — all generated automatically.

Security Assessment Report

app.acme-corp.com · 71 minutes · 12 unique findings

10 Confirmed 1 Claimed 1 Debunked

Quant Sentinel performed an automated security assessment of the ACME Corp web application, analysing both the running application and its source code repository. Of 12 unique findings: 10 were confirmed with machine-verified proof, 1 remains claimed (witness captured, not reproduced), and 1 was actively debunked with the rejecting response preserved — because you should trust the report in both directions.

25+
AI Agents
5+
Pipeline Phases
OWASP
Top 10 Coverage
2–3 hrs
Full Assessment

Five-phase pipeline, plus follow-through

Each assessment follows a structured pipeline that adapts to your target and testing mode — then keeps working what it finds.

1

Scoping

Analyses source code to identify attack surface, tech stack, and high-risk areas.

2

Pre-reconnaissance

Deep code analysis to find potential vulnerability patterns before live testing begins.

3

Reconnaissance

Browser-based mapping of the running application plus ffuf content discovery — any route recon misses is automatically re-tasked via coverage-gap repair.

4

Vulnerability + Exploitation

Parallel specialist agents analyse and exploit. Confirmation is deterministic: an exploited verdict requires machine-checkable proof captured from the target — no self-attested findings.

5

Discovery & Chaining

New vectors found mid-exploitation feed targeted second passes; won sessions and credentials feed chained privilege-escalation testing.

6

Reporting

Executive summary, detailed findings with CVSS scores, deterministic findings roster, and SARIF export for CI/CD integration.

Testing modes

Adapt to Any Engagement

Choose the testing mode that matches your access level. White-box adds deep source code analysis. Grey-box uses API specifications. Black-box relies entirely on dynamic testing — no source or spec required.

  • Pipeline adapts automatically to available information

  • Safe mode skips exploitation for risk-averse scans
  • Scoped analysis via source paths or URL rules
Testing modes
White-box
  • Source code
  • API spec
  • Browser testing
  • HTTP tools
Grey-box
  • Source code
  • API spec
  • Browser testing
  • HTTP tools
Black-box
  • Source code
  • API spec
  • Browser testing
  • HTTP tools
Choose based on your access level and target type

Target types

Web Apps, APIs, or Both

Webapp agents use Playwright for browser-based testing. API agents use HTTP and GraphQL tools. Hybrid mode runs all vulnerability types in parallel — up to 12 specialist tracks covering both attack surfaces simultaneously, including a dedicated GraphQL track and file-upload analysis.

  • Blind vulnerability confirmation via out-of-band (OOB) canary callbacks — blind SSRF, XXE, SSTI, and blind/stored XSS are proven, not just suspected
  • ffuf content discovery finds unlinked endpoints crawlers miss
  • REST, GraphQL, and WebSocket endpoint testing
  • Raw HTTP control — binary bodies, duplicate headers, cookie jars, redirect chains
Target types
Web application
5 vuln types
Playwright browser automation
Injection XSS Auth SSRF Authz
API
8 vuln types
HTTP & GraphQL tools
Injection Auth SSRF Mass assignment API misconfig API resource abuse API consumption Authz
Hybrid
9 vuln types
Browser + HTTP + GraphQL
All webapp + all API vulnerability types
OWASP-aligned vulnerability coverage

Configuration

Authenticate and Scope

Describe login flows in natural language. Support for form-based login, TOTP two-factor, OAuth2, and API keys. Define focus and exclusion rules to target specific areas of your application.

  • Natural language login flow descriptions
  • TOTP 2FA support built-in
  • URL path and subdomain scoping rules
sentinel.auth.yml
login_type : form
credentials :
username : admin@example.com
password : $${SENTINEL_PASSWORD}
totp_secret : $${SENTINEL_TOTP_SECRET}
login_flow : >
Navigate to /login, enter username and password,
enter TOTP code from authenticator app,
click Sign in button.
rules :
focus :
- /api/v1/**
- /dashboard/**
avoid :
- /logout
- /admin/delete-all

Automation

Security in Your Pipeline

Run Sentinel in CI/CD with —quiet mode. Export findings as SARIF 2.1.0 for integration with GitHub Security, GitLab SAST, and other scanning tools. Track costs per scan for budget management.

  • SARIF 2.1.0 with OWASP and CWE mappings
  • Ground-truth detection benchmarking — gate releases on recall and false-positive regression thresholds
  • Docker image for containerised execution
  • Per-agent cost tracking and session metrics
CI/CD Pipeline main branch
Running
Push
Build
Test
Sentinel
Deploy
SARIF output sentinel.sarif
{
"version" : "2.1.0" ,
"runs" : [ { ... } ],
"results" : [
{ "ruleId" : "CWE-89" , "level" : "error" }
]
}
GitHub Actions · GitLab CI · Bitbucket Pipelines View docs

Reporting

Evidence, Not Assumptions

Confirmation is deterministic, not self-attested: an exploited verdict requires machine-readable proof captured from your target. Findings that can’t be proven are debunked with the same rigour — you get evidence both ways, never assumptions. Executive summaries for leadership, detailed technical findings for engineers.

  • Proof-verified exploitation — every Confirmed finding is machine-checked against captured evidence
  • False positives actively debunked with code-level proof, not silently dropped
  • CVSS 3.1 scoring with CWE references
  • Actionable remediation with code examples
INJ-VULN-01

Critical · CVSS 9.8 · proof-verified

SQL Injection in Search Endpoint

CWE-89 · src/pages/api/search.ts:47

// Vulnerable code
prisma.$queryRawUnsafe(
`SELECT * FROM products

WHERE name ILIKE ’%${req.query.q}%’`

)

Exploited: UNION SELECT version() returned PostgreSQL 16.2

Trust layer

A Report You Can Defend

AI-generated findings are worthless if you can’t trust them. Every layer of Sentinel’s output is machine-checked before it reaches you — so Confirmed means proven, and everything else is honestly labelled.

  • Deterministic confirmation — exploited verdicts require a proof token captured verbatim from your target

  • Anti-hallucination grounding — every code citation is verified against your repository; fabrications fail and retry

  • Honest negatives — findings that can’t be proven are debunked with the rejecting evidence preserved

  • Deduplicated, reconciled output — cross-queue findings merge with aliases kept; nothing drops silently

Finding status, determined by evidence
ConfirmedReproduced live and machine-verified against captured response proof
ClaimedConcrete supporting detail, but no proof-verified reproduction this run
UnverifiedTheoretical, or actively disproven — the rejecting evidence is preserved either way

Infrastructure

Beyond the Application Layer

Scan your network infrastructure alongside your applications. Feed a list of IPs, domains, or CIDR ranges and Sentinel maps open ports, fingerprints services, analyses TLS configurations, and matches known CVEs — all in a single automated pipeline.

  • Port scanning, service fingerprinting, and DNS subdomain discovery
  • TLS/SSL analysis — weak ciphers, expired certs, protocol vulnerabilities
  • CVE matching against 8,000+ vulnerability templates
  • Technology detection — CDN, framework, server, and platform identification
  • Automatic webapp candidate identification for follow-up application testing
Infrastructure Pipeline 3 phases · 6 agents
1
Discovery
Port scanning · Service fingerprinting · DNS enumeration
2
Analysis — 4 agents in parallel
TLS/SSL Network Web CVE
3
Reporting
Consolidated assessment · SARIF export · Webapp candidates
Host inventory
portal.example.com
80, 443 · nginx, PHP 8.4
api.example.com
443, 8443 · Express, Node.js
10.0.1.50
22, 445 · OpenSSH, SMB
3 hosts · 7 ports · 6 services 4 findings

Start your first automated pentest

Deploy Sentinel against your application and get a comprehensive security assessment in hours, not weeks.

Full OWASP Top 10 coverage
Application and infrastructure scanning
White-box, grey-box, or black-box
SARIF export for CI/CD