ARMOUR
Sign inGet started →

Know what's vulnerablebefore it's exploited.

Test your website, code, and APIs for real security vulnerabilities before attackers find them.

https://

Scans are charged the input tokens they read, nothing else.

67 checks

JWT weaknesses, Weak or known passwords accepted, Authentication bypass, Brute force and rate limiting, Default credentials, Session fixation, predictability or leakage, Insecure cookie flags, MFA or OTP bypass, IDOR and object-level access, Privilege escalation, Broken function-level authorization, Multi-tenant and cross-tenant access, Hidden, debug or internal endpoints, BOLA and IDOR, Mass assignment, Excessive data exposure, GraphQL introspection, authz and depth, Rate limits and resource consumption, SSRF through the API, Webhook manipulation, auth and replay, Shadow, undocumented or deprecated endpoints, XSS, stored, reflected and DOM, CSRF, Open redirect, Path traversal, LFI and RFI, Malicious file upload and Zip Slip, Insecure deserialization, HTTP request smuggling, Prototype pollution, Race conditions, SQL injection, all forms, NoSQL injection, OS command injection, SSTI, template and expression injection, XXE and XML injection, LDAP and XPath injection, Header and CRLF injection, Remote code execution, Price, currency and discount manipulation, Payment replay, double refund, checkout bypass, Workflow and state machine bypass, Quota, limit and eligibility bypass, Race condition exploitation, Cloud keys: AWS, Azure, GCP, Git and CI tokens, API keys, JWT, encryption and signing secrets, .env, git history and log secrets, Private and SSH keys, Public storage, database or cache, IAM misconfiguration and excessive permissions, Cloud metadata access, Firebase and Supabase RLS or storage rules, Known-vulnerable dependency, Vulnerable version deployed and reachable, Missing or weak HSTS and CSP, Wildcard or credentialed CORS, Stack traces, verbose errors, version disclosure, .git, .env, backup and source map exposure, Weak TLS, cache and host header trust, Open ports and exposed services, Container privileges, docker socket, capabilities, Kubernetes RBAC, pod security, API server, Unsafe workflows, dependency confusion, poisoning, Prompt injection: direct, indirect and RAG, System prompt, secret and cross-user leakage, Tool abuse, excessive agency, SSRF through tools, Insecure output handling and generated injection

Code ships faster than it gets tested.

A year of one team’s work, week by week. Only the red bar was checked for security problems.

250

days something shipped

1

day it was tested

That leaves 249 days of shipped code that nobody looked at.An illustration: one team, one year.

Scan. Prove. Fix. Prove it again.

A finding is not closed when someone says it is. It goes round the same loop until the attack no longer works.

acme-store.com / scanqueued
$ scan acme-store.com
target website · https://acme-store.com
depth standard
checks 67 in 12 areas
cost input tokens read
✓ you confirmed you may test it
▶ scan started

Three ways in. You pick one.

A web address is enough to start. Your source code is optional.

Your site’s address

https://acme-store.comStart scan
acme-store.com

Pages it can reach

  • /
  • /products
  • /cart
  • /login
  • /account

Only what a visitor can already open.

What that catches

It changes a number in the address

Signed in as one customer, it asks for another customer’s invoice.

GET /invoices
GET /invoices/4471 200 yours GET /invoices/4470 200 theirs

What you are handing over

The public surface

Everything you need to test and fix.

Targets, checks, proof, fixes and reports, from one account.

The checks

67 checks across 12 areas

From sign-in and access control to injection, secrets and cloud. Run them all, or pick the areas you care about and how many passes to make.

Authentication

8 checks

Authorization

5 checks

API security

8 checks

OWASP web

9 checks

Injection

8 checks

Business logic

5 checks

Secrets

5 checks

Cloud

4 checks

Dependencies

2 checks

Headers and configuration

5 checks

Infrastructure

4 checks

AI and LLM

4 checks

Live results

Findings land as they are found

Watch the scan work. Each finding is filed the moment it is proved, so you can start reading before it finishes.

Scanning acme/store· 41 of 67 checks

criticalSQL injection in the sign-in formapp/api/login/route.ts:14
highAnother customer’s invoice is readableapp/api/invoices/route.ts:9
mediumThe session cookie has no Secure flaglib/session.ts:31

Pull requests

Every pull request, read before it merges

Connect GitHub and each pull request gets a review of what it changes, as a comment on the pull request itself.

Armour reviewed this pull request2 findings

fix: look invoices up by account #214

highAnother customer’s invoice is readableapp/api/invoices/route.ts:9
mediumThe session cookie has no Secure flaglib/session.ts:31

Attack surface

See what is exposed, and the way in

Hosts, open ports, services and the paths between them, mapped from the scan, with the route an attacker would take marked.

:443 https:22 ssh/admin/api/v1db.internal:8080 stagingacme-store.comhost

Reports

A written report, ready to hand over

Every scan makes a report you can print or save as a PDF, with the raw files beside it.

Written reportPDFFindings as JSON

Tokens

You pay for what it reads

Each plan gives you input tokens every month. A scan is charged what its model reads, and seats are free.

Your data

Nothing kept that you did not mean to

Your code is used for the scan and removed when it finishes.

  • Only targets you confirm you may test
  • Every scan is logged
  • Erase your account whenever you like

Not a scanner. Not a pen test.

It attacks the way a person would, shows the proof, and writes the fix.

01

Proof, not a list

Every finding keeps the request it sent and what came back, so you can see it happen for yourself.

02

A fix, not just a PDF

Each finding says what to change. For code, it writes the patch, and then checks that the fix held.

03

When you need it

Run it when you choose, or on every pull request. You pay for the input tokens it reads, nothing else.

A scanner that lists
An annual pen test
Armour
Shows the proof behind each finding
—
✓
✓
Writes the fix for you
—
—
✓
Checks that the fix held
—
Extra
✓
Runs whenever you need it
✓
—
✓
Charges only for what it uses
Varies
A quote
✓

Typical tools and services of each kind, not a review of any one of them.

From broken to protected in one click.

Built your site with AI? Drop the code in. We read every line, explain what's wrong, and fix it for you.

Sign in. Continue with Google or GitHub, or use an email and password. No separate sign-up.

A sample project: yours will have its own findings.

One issue. One fix.

Each finding shows what was tried, what came back, and the fix.

HighCWE-639OWASP API1GET /api/v1/invoices/{id}

Any signed in customer can read another customer’s invoices.

  1. 01  What it tried

    GET /api/v1/invoices/4471Signed in as an ordinary customer. Ours.

    GET /api/v1/invoices/4470Same login. One number changed.

    HTTP/1.1 200 OK
    { "customer": "Brightline Ltd",
      "card":     "**** 4417" }
  2. 02  Why it worked

    app/api/invoices/route.ts

    8const { id } = req.query9const invoice = await db.invoice.findUnique({10  where: { id },11})12return Response.json(invoice)

    The lookup never checks who is asking. Any signed-in customer can name any invoice.

  3. 03  The change

    app/api/invoices/route.ts

    - const invoice = await db.invoice.findUnique({-   where: { id },+ const invoice = await db.invoice.findFirst({+   where: { id, accountId: session.accountId },  })

    Now it only finds the signed-in customer’s own invoices.

✓ FixedAsked again: GET /api/v1/invoices/4470 → 404 Not FoundRe-verified after the change.

Made for whoever ships the code.

From one developer with a weekend project to a company whose customers ask for the report.

The questions everyone asks first.

Plain answers, before you run anything.

Something else? Ask us
Will it break my site?No

It never deletes data, changes settings or places orders. Where a test could do damage, it stops and reports what it would have done.

Can I test something I do not own?No

You confirm you own the target, or may test it, before a scan starts.

What happens to my code?Removed

It is used for the scan and deleted when the scan finishes.

Can I see what it tested?Yes

Every request it sent, and what came back, is in the report.

You pay for what it runs.

Each plan gives you input tokens every month. A scan is charged the tokens its model reads, after it reads them, and nothing else.

StarterRecommended

$20/ month

One small site, checked before you ship.

1.5Binput tokens a month

About 5 completed scans

  • Up to 5 targets
  • A web address only
  • The full report, with every fix
Start on Starter
Business

$500/ month

Several services, and customers asking for the report.

60Binput tokens a month

About 200 completed scans

  • Up to 80 targets
  • A web address only
  • The full report, with every fix
  • Hand over an API
Start on Business
Enterprise

Talk to us

Everything you own, source included.

Tokens to suit
  • As many targets as you need
  • A web address only
  • The full report, with every fix
  • Hand over an API
Book a call
  • Free to startSign up free, then top up
  • A completed scan150M–310M input tokens
  • Top-ups$20 per billion tokens, never expire
  • PricesUS dollars, excluding tax. Seats are free

Find out what an attacker would find.

Enter your target. You pay only for the tokens the scan reads.

https://

Or scan an API instead, or review your code.

Nothing to install

It runs on our machines. Nothing goes near your servers.

Charged for what it reads

A scan is charged the input tokens its model reads, after it reads them.

$20 to start

Starter is one small site, checked before you ship.