Automated Website Vulnerability Scanner
Paste the URL of any website. The scanner loads it in a real browser, runs the checks, and returns findings with evidence and a severity for each one.
SCAN YOUR WEBSITE NOW
Enter the URL of a site you own or have written permission to test. The scan starts from the public surface, the same view an attacker has.
An automated website vulnerability scanner tests a live website from the outside and reports weak points with proof. It works on any site, not only AI-built apps. Paste a URL, wait for the scan, then fix the findings in order of severity.
What is an automated website vulnerability scanner?
An automated website vulnerability scanner is a program that attacks your site in a safe way and records what happens. It loads the site, follows the links, and sends test requests. It compares each response to patterns of known weakness.
This type of tool is also called an online vulnerability scanner, a web vulnerability scanner, or a black-box DAST scanner. DAST means dynamic application security testing. “Black-box” means the tool does not see your source code. It sees only what a visitor sees.
The term applies to any website. A WordPress blog, a shop on a hosted platform, a React single-page app, and a server-rendered portal all have the same public surface. Each one has headers, TLS settings, files, scripts, and endpoints that the scanner can examine.
If your site was built with an AI coding tool, read the vibe code scanner page. It covers the failure modes that are common in those apps. This page covers the general case.
How do I scan a website for vulnerabilities?
Use this procedure for the first scan.
- Make sure you own the site or have written permission to test it.
- Paste the public URL into the scanner. Use the production address, not a local copy.
- Choose a profile. Use Quick Scan for a first look, or DAST Full for a deep pass.
- If the site has a login, add a credential in the Options step. Use a test account, not an admin account.
- Add excluded paths for routes that must not be touched, such as
/logout. - Start the scan and wait for the status to change to completed.
- Read the findings from the highest severity to the lowest.
- Fix each finding, deploy, and run a retest.
The scan profiles in VibeEval differ in depth and time.
| Profile | Typical time | What it does |
|---|---|---|
| Quick Scan | 1-3 minutes | Passive checks, header analysis, and common vulnerabilities |
| DAST Full | 10-30 minutes | Deep crawl, injection tests, and fuzzing |
| API Scan | 5-15 minutes | Tests REST and GraphQL endpoints from an OpenAPI spec |
| Custom | Varies | You pick individual test modules |
The full list of options is on the Running scans page. The rules for valid URLs are on the Targets page.
What does an automated scanner check?
A good scanner covers several layers. The list below follows the check modules in the security checks reference.
| Area | Example checks |
|---|---|
| Security headers | Missing or weak Content-Security-Policy, short HSTS max-age, CORS that reflects any origin |
| TLS and transport | TLS configuration, mixed content, forms that post over HTTP |
| Exposed files | .git folders, .env files, backup files, source maps, directory listings |
| Secrets in the client | API keys in scripts and bundles, tokens in localStorage, tokens in URLs |
| Access control | IDOR, privileged fields in responses, mass assignment, authorization bypass |
| Sessions and tokens | JWT with no expiry, weak algorithms, missing CSRF tokens |
| API faults | GraphQL introspection, REST endpoints tested for IDOR |
| Injection | SQL, command, and template injection, DOM XSS, CRLF injection |
| Infrastructure | Open redirects, path traversal, SSRF, subdomain takeover, outdated libraries |
Headers, TLS, and exposed files
Headers are the easiest area to test. The scanner reads the response headers and compares them to a safe baseline. A missing Content-Security-Policy is a common result. See the MDN guide to Content Security Policy for the fix.
The scanner also asks for files that should not be public, such as .git directories and .env files. If the server returns the file, the finding shows the request and the response as evidence.
Keys in bundles
Many sites ship JavaScript bundles with keys inside them. The scanner reads inline scripts and bundles and looks for provider keys. It checks for keys from AWS, OpenAI, Stripe, GitHub, Slack, Supabase, Firebase, and others.
If you find a leaked key, rotate it first. A deploy that removes the key from the bundle does not make the old key safe. For a focused check, use the Token Leak Checker.
Authentication and access control
If you give the scanner a credential, it can test pages behind the login. It attaches the credential to every request. The scanner then tests for broken object-level authorization, privileged fields in responses, and mass assignment.
If the site uses Supabase, the scanner also checks for tables that are readable without Row Level Security. For a focused check, use the Supabase RLS Checker.
APIs and injection
If you give the scanner an OpenAPI or Swagger URL, the API profile knows every endpoint and parameter. It then tests each one for injection, IDOR, and mass assignment.
For injection, the scanner sends test payloads in query strings, form fields, and headers. It looks for SQL errors, reflected script, and command output in the responses. The OWASP Foundation describes these classes in the OWASP Top 10.
What can an automated scanner not check?
Every scanner has limits. Know them before you trust a clean result.
- Source code. A black-box scanner does not read your code. It cannot find a hidden backdoor or an unsafe function that no request reaches.
- Internal networks. The scanner tests public URLs only. It cannot test a private admin network.
- Business logic. A scanner does not know your rules. It cannot tell that a discount code should work only once.
- Pages it cannot reach. If a page needs a login that you did not provide, the scanner does not see it.
- Unknown hosts. The scanner tests the URL you give. A forgotten subdomain stays untested until you add it. See discover forgotten web apps.
VibeEval lists these limits in each report. The on-call note has a section named “Out of scope” that shows what was not covered in that pass. The methodology page gives more detail on black-box limits.
For the difference between a scanner and a deeper test, read vulnerability scanning vs AI pentest. If you need a signed report for an auditor, see pentest services.
How do I read the scan results?
Each finding has the same fields. Learn them once and you can read any report.
| Field | What it tells you |
|---|---|
| Title | A short name, for example “Supabase table exposed without RLS” |
| Severity | Critical, high, medium, low, or info |
| CVSS | A base score from 0 to 10 |
| CWE and OWASP | The weakness class and the Top 10 category |
| Evidence | The HTTP request and response that showed the issue |
| Remediation | Numbered steps to fix the issue |
| Status | Open, confirmed, false positive, or fixed |
Use the severity to decide the order of work.
- Fix every critical finding first. Examples are secret keys in a bundle and tables that anyone can read.
- Fix high findings next. Examples are access control faults and open admin routes.
- Plan medium findings for the next sprint. Missing headers and open redirects are common here.
- Review low and info items when you have time.
A medium finding can become a high finding when it is chained with another one, for example an open redirect plus a token in a URL.
The scan page also shows a risk score from 0 to 100. The risk score page explains the formula. Use the score to compare one scan to the next, not to decide that the site is safe.
If a finding looks wrong, mark it as a false positive. Use the evidence to decide. Read the findings reference for all actions on a finding.
Is it safe to run a scanner on a production website?
A scan of a production site is safe when you control the settings. VibeEval scans are production-safe by default. They probe, but they do not destroy data.
Use these settings to lower the risk further.
- Tick Passive only for a read-only pass. The scanner observes and never injects.
- Lower Max requests per second. The default is 50.
- Add Excluded paths for destructive or noisy routes, such as
/logoutand/admin/*. - Use a test account for authenticated scans. Do not use a real customer account.
- Add a Custom header to tag the scanner traffic in your logs.
- Tell your hosting provider and your security team before the first scan.
A contact form can send many emails during a scan. A signup form can create many accounts. Exclude these routes.
Staging copies have a limit. VibeEval refuses localhost, private addresses, and hosts that end in .local, .internal, or .test. To test a pre-production build, deploy it to a public preview URL and scan that.
Who is allowed to scan a website?
Scan only sites that you own or have written permission to test.
VibeEval does not ask for a DNS record or a verification file. It relies on a safety filter and on abuse review. It refuses government and military domains. It suspends accounts that scan third-party systems.
If you scan a client site, get a written scope letter first. The letter must list the domains, the dates, and the contact person. See multi-site security monitoring for the work of many client sites.
How often should I scan a website?
Scan after each change that can affect the public surface. These changes include a new deploy, a new integration, a new subdomain, and a change to the DNS or the CDN.
A single scan shows the state at one time. A new route, a new key, or a new table can add a new weakness later.
Use Re-run Scan to repeat a scan on the same target with the same profile. Each re-run produces a trend that shows which findings are new, fixed, or unchanged. Findings are deduplicated by target and title, so a re-run does not create a second copy of an issue.
For always-on checks, read security monitoring. Monitoring is an early-access capability in VibeEval, so treat re-runs as the dependable path today. For a check on each deploy, read continuous security smoke testing. For depth on each release, read continuous penetration testing.
How do I verify a fix?
A finding is not closed until a retest proves the fix.
- Apply the fix in the code or in the host settings.
- Deploy the fix to the same environment that you scanned.
- Use Retest on the finding. This re-runs the single check that produced it.
- Make sure the finding changes to fixed.
- Run a full re-run to make sure no new finding appeared.
- For a leaked key, confirm that the old key no longer works.
The Copy for LLM option exports the whole report as Markdown. You can paste it into a coding agent. Each finding includes the evidence, the fix steps, and a rule that tells the agent to make the smallest fix possible. Review the patch before you deploy it.
What else should I use with a scanner?
A scanner is one layer of a security program. It does not replace the others.
| Layer | What it adds |
|---|---|
| Code review | Finds logic faults and unsafe code before release |
| SAST and dependency checks | Find unsafe code patterns and known CVEs in the source |
| Automated website scanner | Finds live faults in the deployed site |
| Human pentest | Tests business logic and chained attacks |
| Monitoring | Finds changes after the deploy |
To compare tools before you choose one, read vulnerability scanner comparison. For the difference between classic DAST and agentic testing, read DAST vs agentic security testing. For a list of tasks for the developer, use the security audit checklist.
NIST also gives a general method for security testing in SP 800-115. The method covers planning, discovery, attack, and reporting. An automated scanner automates the discovery and the attack steps.
What are the common mistakes?
- Scanning only the marketing domain. The app often lives on a subdomain. Scan each host.
- Ignoring medium findings. Attackers chain them.
- Not rotating a leaked key. A rescan does not make the old key safe.
- Treating one clean scan as permanent. The next deploy can open a new weakness.
- Opening tickets on preliminary findings. Wait for the on-call review to finish. The findings can change.
Use the scanner to find the obvious faults first. Then use deeper tests for the faults that a scanner cannot see.
Common questions
What is an automated website vulnerability scanner?
Is an online vulnerability scanner safe to run on a production website?
Do I need to install anything or give source code access?
Can the scanner test websites that are not AI-built?
Does a clean scan mean the website is secure?
Can I scan a website that I do not own?
Test your website today
Start a 14-day trial with no credit card. Run a Quick Scan in minutes or a full DAST scan, then have an on-call security researcher confirm each finding.
14-day free trial · Cancel anytime