Skip to main content
Cloudflare Launchpad About the Launchpad

Penetration testing: prove the attack,
keep the evidence.

ShieldProbe Assess pentests your web apps, APIs, networks and servers, and you buy it as Penetration Testing as a Service: scoped tests, findings and retests in one portal. Our AI runs the attack, and a named consultant approves each finding before you see it. With scope agreed and access ready, the report can arrive in 48 hours. New to pentesting? See how a penetration test works.

Takes a minute. We contact you to agree targets, access and timing.

Follow one finding through Assess

Example Store, Orders API, with fictional data. Testing networks or servers? See what we test.

Choose a step to see what Assess records at each stage.

SP-014 Evidence record

The same request.
A different customer.

Signed in as customer-aOrder belongs to customer-b
GET /api/orders/order-b07

Response not yet replayed

A valid login should not grant access to someone else's order. ShieldProbe tests the authorization boundary.

The Assess portal, shown with fictional data.

Read the evidence
SP-014Illustrative API case

Assess / The exploit

Same requestGET /api/orders/order-b07
Signed in as
customer-a
Order belongs to
customer-b
200 OK
{
  "id": "order-b07",
  "ownerId": "customer-b"
}

Another customer’s order was returned. The ownership check is missing.

Consultant review pending. The request, response and customer context go to the reviewer together.

What we test

Web, API, network and server pentesting.

Choose the systems to test, where they run and the access we start with. Each assessment has an agreed scope and consultant-reviewed evidence.

Web applications

Customer portals, dashboards and browser-based applications.

Test authentication, sessions, access controls and business workflows. Follow weaknesses through to their demonstrated impact.

Location
Internal or external
Access
Black-box or grey-box
Methodology
Agreed OWASP Top 10 or full WSTG scope

APIs

REST, GraphQL and SOAP APIs, including backend services and integrations.

Test endpoint authorization, data exposure, input handling and API workflows, using the agreed accounts and API specification.

Location
Internal or external
Access
Black-box or grey-box
Methodology
Agreed OWASP API Security Top 10 or full WSTG scope

Networks & server infrastructure

Internet-facing services, private networks, and on-premises or cloud-hosted servers.

Test exposed services, misconfigurations and exploitable weaknesses. Validate privilege escalation and lateral movement where the agreed scope permits.

Location
Internal or external
Access
Black-box only
Methodology
Agreed PTES scope
Internal or external?

External: test internet-facing systems from our infrastructure.

Internal: reach private applications, APIs, networks and servers through an approved connector or access path.

Black-box or grey-box?

Black-box: start without supplied login credentials and test the reachable, authorized surface.

Grey-box: use supplied test accounts and context to test authenticated web and API workflows.

Network and server infrastructure assessments currently use black-box testing.

Not sure where to start? We'll help you define targets, access and permitted activities before testing begins.

Scope your assessment

A valid login.
An invalid permission.

Requesting account
customer-a
Returned order belongs to
customer-b
GET /api/orders/order-b07
Authorization: session of customer-a

HTTP/1.1 200 OK
{ "id": "order-b07", "ownerId": "customer-b" }

The server returned another customer's order. Fix the ownership check in code. Defend can log this request but cannot block it, because a rule cannot see who owns an order.

The deliverable

Evidence your team can act on.

An Assess report connects tested scope, reproducible findings, business impact and a clear next action.

This example uses fictional data to show how a report is laid out.

Scope your assessment
Illustrative sample reportShieldProbe Assess

SP-014 High · illustrative severity

Order accessible by another customer

Scope
Example Store, Orders API. Two authorized test accounts.
Observed impact
Customer A can retrieve Customer B's order.
Required correction
Check object ownership in application code.
Read sample report

Executive summary and scope

A fictional grey-box API test, not a completed customer engagement: one authenticated route, tested with two test users. A real report sets out everything in scope.

Method and reproduction

Sign in as customer-a. Request order-b07, which belongs to customer-b. Compare the returned owner with the authenticated account.

GET /api/orders/order-b07
Authenticated account: customer-a

HTTP/1.1 200 OK
{ "id": "order-b07", "ownerId": "customer-b" }

Impact and severity rationale

One customer can read another customer's order: unauthorized disclosure. Changing or deleting the order was not shown. High assumes the order data is sensitive; a real report gives the severity method and business context behind each rating.

Remediation and protection

Bind the lookup to the signed-in owner and review the intended shared-team access policy. Defend can log this request but not block it, because a rule cannot see sessions or the database to decide who owns an order.

Illustrative verification record

Patched staging, Defend off: cross-customer access returns 404, the legitimate owner returns 200, and a signed-out request returns 401. Shared-team access still needs a policy decision. Sample results.

Review and delivery

A named consultant approves every Assess finding before delivery. Your engagement sets the authorization, targets, method, access, exclusions, delivery date and retest terms. The report supports your audit; acceptance is your auditor's decision.

How an Assess pentest runs

  1. Scope

    We agree targets, written authorization and testing windows with you.

  2. Test

    Our AI plans and runs the attack from our own infrastructure. Internal targets are reached through an approved connector or access path.

  3. Review

    The AI authors each finding. A named consultant approves it before you see it, and the gate is enforced in the database.

  4. Report

    Reproducible request and response evidence. The 48-hour clock starts when scope, authorization and access are ready; larger scopes may take longer.

  5. Retest

    Retest fixed findings for 12 months, under your engagement terms.

Bring to scoping

  • Target owners, roles and written authorization
  • Test accounts, API specifications and agreed data
  • Verified internal access or allowlisting
  • Testing windows, stop conditions and escalation contacts
  • The evidence you need, delivery date and retest terms
How does ManticoreAI make sure AI-driven testing doesn't damage systems?

Before anything runs, we agree the targets, access, testing windows, rules of engagement and an operational contact with you. Testing stays inside that scope. Any pentest carries some risk of disruption, so for sensitive systems we can test a staging environment or agree a quiet window.

Can AI replace human experts in penetration testing?

No, and we do not run it that way. The AI authors the finding; a consultant reviews and approves it before you see it. We call that Auditor, not Author.

What makes the report audit-grade?

The report includes the agreed scope, reproducible exploit evidence, business impact, and remediation guidance. Findings are reviewed and approved by a named consultant before delivery. Your auditor can follow what was tested and what the evidence demonstrates.

How long does it take from first call to report?

The 48-hour delivery applies once scope is agreed and access is ready. Before that, we agree targets, authorization and testing windows with you. Larger scopes, unavailable targets or further investigation can take longer.

What types of penetration testing does ManticoreAI offer?

ShieldProbe Assess tests web applications, REST/GraphQL/SOAP APIs, networks and server infrastructure. Assessments can cover internal or external targets. Web and API assessments support black-box or authenticated grey-box testing; network and server infrastructure assessments currently use black-box testing. We agree the targets, access and permitted activities with you before testing starts.

Can you test internal systems as well as public targets?

Yes. External testing reaches internet-facing applications, APIs and services from ManticoreAI infrastructure. Internal testing reaches private applications, APIs, networks and servers through an approved connector or access path. Location and access mode are separate: an internal target may still be tested without login credentials.

What is the difference between black-box and grey-box testing?

Black-box testing starts without supplied login credentials and tests what an attacker can discover and reach within the authorized scope. Grey-box testing uses supplied test accounts and context to examine authenticated web or API workflows and permission boundaries. Web and API assessments support both modes. Network and server infrastructure assessments currently support black-box only.

Leave with evidence you can act on.

Scope your assessment