Pentest your web app with Claude

Point it at any web app. The crawler maps every page, form, and endpoint. Claude examines each one like a senior pentester and delivers findings with reproduction steps and remediation advice.

3-day free trial, no credit card required

scan - live output

Scan started for https://example.com

Using model: claude-sonnet-4-6

Crawling target (depth=3, rate=2 req/s)

Crawl complete: 24 pages discovered

Analysing page 1/24: /login

Found 2 vulnerabilities on /login

· Missing CSRF token on login form [high]

· Password field with autocomplete enabled [medium]

Analysing page 2/24: /api/users

Found 1 vulnerability on /api/users

· IDOR: numeric user IDs in path, no authz check [critical]

Analysing page 3/24: /admin

Active probing found 1 confirmed vulnerability

· XSS: payload reflected in search results [high]

Analysing page 4/24

How it works

01

Enter a URL

Paste any URL: public site, staging environment, or internal app. Configure depth, rate limits, and authentication if needed.

02

Crawl & analyse

The crawler maps every page, form, and endpoint. Claude examines each one for vulnerabilities, considering full page context — headers, cookies, HTML, API responses.

03

Get actionable findings

Each finding includes a severity score, exact reproduction steps, and remediation guidance.

What makes it different

Signature-based scanners miss logic flaws, auth bypasses, and novel attack patterns. AI reasoning finds what rules can't.

Reasons over context, not patterns

Claude reads HTML structure, HTTP headers, cookies, forms, and API responses together — finding logic flaws and auth bypasses that signature-based scanners miss.

Crawls the whole site

Playwright-powered crawler discovers every page, form, endpoint, and SPA route. Up to 50 pages per scan with configurable depth.

Confirms with active probing

Optional fuzzing injects XSS, SQLi, path traversal, and open redirect payloads to confirm exploitability — not just flag suspicion.

Scans behind logins

Provide credentials once and the crawler handles session management to test your authenticated attack surface, including admin panels and protected APIs.

Streams results live

Every page analysed, every finding surfaced, and cost tracked — streamed to your dashboard as the scan runs.

Exports to PDF

Reports include severity ratings, reproduction steps, and concrete remediation advice for every finding.

What it looks for

16 vulnerability categories across the OWASP Top 10 and beyond.

XSSCross-site scripting
SQLiSQL injection
IDORBroken object-level auth
CSRFCross-site request forgery
SSRFServer-side request forgery
Broken authAuth & session flaws
Open redirectUnvalidated redirects
JWT issuesalg:none, weak secrets
CORSMisconfigured origins
Info disclosureKeys & secrets in HTML
Security headersMissing CSP, HSTS, etc.
Insecure cookiesMissing HttpOnly/Secure
Mass assignmentOver-posted fields
Path traversalDirectory escape
Exposed admin/admin, /actuator
Verbose errorsStack traces & debug info

Findings you can act on

Every vulnerability includes evidence, reproduction steps, and a concrete fix.

CriticalIDOR: Unauthenticated Access to User Records
IDOR

Description

The /api/users/[id] endpoint returns full user records including email, address, and payment info without verifying the requesting user owns the resource. Sequential integer IDs make mass enumeration trivial.

Reproduction steps

  1. Authenticate as any user
  2. Note your user ID from session
  3. Request /api/users/1, /api/users/2, …
  4. Observe full PII returned for all IDs

Remediation

Enforce ownership check server-side: compare the requested resource's owner ID against the authenticated session user before returning data. Use opaque UUIDs instead of sequential integers.

Confidence
0.92
https://example.com/api/users/[id]

Simple, flat-fee pricing

Bring your own Anthropic API key and pay a single platform fee. No hidden charges — your AI costs go directly to Anthropic.

Start free trial3-day free trial, no credit card required
$9/mo

Connect your Anthropic API key. You pay Anthropic directly for scan usage — the $9 covers the platform.

  • Full platform access
  • All models — Haiku, Sonnet, Opus
  • Authenticated scanning
  • Scheduled scans
  • PDF reports with remediation guidance
  • Real-time scan progress stream
Start free trial

3-day free trial, no credit card required

Frequently asked questions

What web vulnerabilities can AI Vulnerability Scanner detect?

AI Vulnerability Scanner detects 16+ vulnerability categories including XSS (cross-site scripting), SQL injection, IDOR (broken object-level authorisation), CSRF, SSRF, broken authentication, JWT issues, CORS misconfigurations, missing security headers, insecure cookies, path traversal, exposed admin endpoints, mass assignment, verbose errors, open redirects, and information disclosure.

How does AI-powered vulnerability scanning differ from traditional scanners like OWASP ZAP or Burp Suite?

Traditional vulnerability scanners rely on signature databases and pattern matching, which miss logic flaws, authorisation bypasses, and novel attack patterns. AI Vulnerability Scanner uses Claude AI to reason over the full context of each page - HTML structure, HTTP headers, cookies, forms, and API responses - identifying security issues that rule-based tools cannot detect.

How long does a web vulnerability scan take?

Most scans complete in 2–10 minutes depending on site size and selected model. A Haiku-powered scan of a 10-page site typically takes 2–3 minutes. A deep Sonnet analysis of a 50-page application may take up to 10 minutes. All scans stream live results to your dashboard as findings are discovered.

Can AI Vulnerability Scanner test authenticated areas behind a login?

Yes. AI Vulnerability Scanner supports authenticated scanning. Provide your credentials once and the Playwright-based crawler handles session management automatically, scanning your entire authenticated attack surface - including user dashboards, admin panels, and protected API endpoints.

Is it safe to run a vulnerability scan on a live production website?

AI Vulnerability Scanner is designed for authorised security testing with configurable rate limits (default 2 requests per second) to minimise impact on production systems. Active probing can be disabled for a passive read-only analysis. Always ensure you have explicit written permission from the site owner before scanning any target.

What is the BYOK (bring your own key) plan?

The BYOK plan at $9/month lets you connect your own Anthropic API key to AI Vulnerability Scanner. Your scan costs are billed directly by Anthropic to your account. You only pay AI Vulnerability Scanner the $9 platform fee. This is ideal if you already have an Anthropic API subscription or want full cost transparency.

Try it on your own app

Paste a URL, pick a model, and you'll have a full vulnerability report in under 10 minutes.

Start your free trial

3-day free trial, no credit card required