Short answer
Penetration testing as a service (PTaaS) is a penetration test you order, authorize, track and retest through an online platform. You scope the target and sign the rules of engagement in the platform, findings appear there as testers confirm them, your engineers fix them, and you request the retest from the same place. The testing is still a penetration test. PTaaS changes how it is bought and delivered.
NIST defines penetration testing as security testing in which evaluators “mimic real-world attacks in an attempt to identify ways to circumvent the security features of an application, system, or network”. A traditional engagement wraps that work in a consulting project: a statement of work, a start date weeks out, a test window, and a PDF at the end. Fixes happen after the testers have left, and checking them is often a second booking.
PTaaS replaces the project wrapper with software. The vendors that rank for this term describe it the same way: HackerOne calls PTaaS “a delivery platform” and Synack says it is “a way to deliver penetration testing, not a separate testing technique”. What differs between providers is who does the testing, how findings are confirmed, how retests are priced, and what the report proves. Those differences are what the rest of this guide is about.
Helix Cyber sells penetration testing as a service, so we have a stake in this category. Where we describe our own offer we say so, and every competitor figure links to the vendor's own page.
How a PTaaS engagement runs
Most PTaaS platforms follow the same six steps. Authorization comes before any testing, and the retest is part of the same engagement.
- 01
Scope
List the domains, URLs, APIs, IP ranges or repositories, pick black, gray or white box, and add test accounts. Good platforms check you own each asset.
- 02
Authorize
Sign rules of engagement that name the exact assets, exclusions and test window. No traffic should be sent before this exists.
- 03
Test
People, AI agents, or both attack the scope. Methodology usually follows OWASP testing guides for apps and APIs and PTES or NIST SP 800-115 for networks.
- 04
Findings as they land
Each confirmed finding appears in the platform with evidence, severity and a fix, instead of arriving weeks later in a PDF.
- 05
Fix and retest
Engineers fix, mark the finding ready, and the provider attacks it again. The retest result is recorded against the original finding.
- 06
Report and letter
A final report and, if you need one, an attestation letter for auditors and customers, both reflecting the retest status.
Step 4 is the change engineers notice first. A critical access control bug found on day one of a two-week consultancy test usually waits for the report. On a platform it can be fixed while the test is still running.
PTaaS vs a consultancy pentest, a vulnerability scanner and bug bounty
These four get compared because they all produce lists of vulnerabilities. They answer different questions. A scanner asks whether known weaknesses are present. A penetration test asks what an attacker can reach, inside a defined scope. A bug bounty pays outsiders for whatever they find.
| Attribute | PTaaS | Consultancy pentest | Vulnerability scanner | Bug bounty |
|---|---|---|---|---|
| What you buy | A scoped pentest, delivered and retested through a platform | A scoped pentest delivered as a project, ending in a report | Software that checks for known weaknesses, usually by subscription | Payouts for valid findings from outside researchers |
| Who tests | Provider staff, vetted freelancers, AI agents, or a mix | The firm's own testers | No one. Automated checks only | Whoever joins the program |
| Is impact proven? | Yes. Findings are exploited within the agreed rules and evidenced | Yes, same standard | Usually not. Matches signatures and versions, so false positives need triage | Yes for each submission, but nothing tells you what was not tested |
| Coverage | Defined scope and methodology, stated in the report | Defined scope and methodology | Broad but shallow. Weak on access control and business logic | Whatever researchers choose to look at |
| When results arrive | As each finding is confirmed | At the end, in the report | Minutes to hours after each scan | Whenever someone submits |
| Retesting | Requested in the platform. Terms range from one retest to unlimited | Often a separate booking or fee | Run the scan again | Researcher may re-check, outside any methodology |
| Pricing basis | Per test, credits, or annual subscription | Day rate times days, quoted per project | Per target or asset, per month or year | Bounty per valid finding, plus platform fees |
| Audit evidence | Yes, if the report shows scope, method, evidence and retests | Yes, the traditional form | Separate control. PCI DSS treats scans and pentests as different requirements | Weak on its own. No coverage statement or methodology |
| Best for | Teams that ship often and need tests, fixes and retests in one loop | Bespoke one-off work: red teams, physical, on-site | Continuous hygiene between pentests | Mature programs that already pentest and want more eyes |
PTaaS vs a traditional penetration test
The test can be identical. The differences are around it: a platform order instead of a statement of work, findings delivered live, retests requested in the same system, and a history of every test in one place. A consultancy still makes sense for work a platform does not cover well, such as a multi-week red team exercise, physical intrusion testing, or a scope that needs testers on site.
PTaaS vs a vulnerability scanner
A scanner compares what it sees against known signatures and versions. It is fast and cheap enough to run every day, and it misses the bugs that need context: one customer reading another customer's invoices, a refund flow that can be replayed, a role that can promote itself. Those are the access control and business logic findings that dominate real pentest reports. PCI DSS draws the same line: it requires quarterly scans and annual penetration tests as separate requirements, and a scan does not satisfy the pentest requirement (see our PCI DSS penetration testing guide). Run both.
PTaaS vs bug bounty
Bugcrowd describes bug bounties as a “pay for results” model in which valid, non-duplicate, in-scope findings are paid from a bounty pool. You pay only for what is found, but nobody commits to covering anything, and the program produces no methodology or coverage statement. Bugcrowd's own glossary notes that pentests are “methodology driven and are best for coverage”. Bounties work best on top of regular testing, once the obvious bugs are gone. Launching one before a first pentest tends to pay outsiders for findings a single test would have caught.
Delivery and retesting models
“PTaaS” covers several ways of scheduling and staffing the work. Ask which one a provider means before comparing prices.
When testing happens
Point in time
One test of a defined scope, ordered when you need it: before a launch, an audit, or a customer security review. The most common way teams start.
Scheduled
Tests on a cadence, such as quarterly or after each major release. Suits teams with a compliance calendar and a steady release rhythm.
Continuous
Ongoing testing of a fixed asset list, often paired with a scanner to flag changes. Check what “continuous” means in hours of human or agent testing per month.
Change-triggered
A test of the affected area when a significant change ships. PCI DSS asks for exactly this after significant infrastructure or application changes.
Who does the testing
- A dedicated team. The same testers each time, who learn your application. Closest to a consultancy relationship.
- A crowd of vetted testers. HackerOne notes that some PTaaS vendors assign a different tester every time. Fresh eyes, less continuity.
- AI agents with human oversight. Agents run the test; people set scope, review and handle what the agents cannot. Faster and cheaper per test, with limits covered below.
How retests work
Retesting is where PTaaS offers differ most, and where a cheap test can become an expensive one. The published terms fall into three groups:
Unlimited during the contract. Cobalt states its PTaaS model “provides unlimited on-demand retesting throughout your contract term”. Helix includes free unlimited retests with every test.
A fixed number per package. BreachLock's packages list 1, 2 or a custom number of free manual retests.
Priced separately. Common in consultancy engagements. Ask for the retest price in the original quote, because PCI DSS requirement 11.4.4 expects exploitable findings to be retested after they are fixed.
How PTaaS is priced
Four pricing models cover almost every PTaaS offer. The figures below are taken from each vendor's public pricing page on 5 November 2026. They are list prices for different scopes and testing depths, so they do not compare like for like.
| Model | How it works | Published examples |
|---|---|---|
| Fixed price per test | A set price for a defined scope, usually one application or API. Easy to compare and budget. | Intruder from $3,500 per AI-powered web app test. Cobalt $3,500 per autonomous pentest, a limited offer running to 31 December 2026. Helix black box $3,000, gray box $4,000, white box from $7,000. |
| Credits or testing capacity | You buy a pool of testing time up front and spend it on tests over the contract. Unused credit can expire. | Cobalt credits equal 8 hours of testing each, sold in annual packages, and do not roll over. Synack lists tiers from $5,085, $10,283 and $27,120, uses credits that expire a year after purchase, and charges for its platform as a separate line item. |
| Annual subscription | A yearly fee for a number of tests or continuous coverage of a set of assets, often bundled with a scanner. | Astra lists Pentest Auto at $2,999 a year, Pentest Expert at $5,999 a year and Enterprise from $9,999 a year. |
| Quote per scope | The provider sizes the work after a scoping call. Common for larger estates and white-box work. | BreachLock quotes its Standard, Extended and Extensive packages, which include 1, 2 or a custom number of manual retests. Cobalt's main tiers are also quote-only. |
For context outside PTaaS: scanners are priced per target, and Intruder charges “a base fee plus a small fee-per-target”. Bug bounties pay per valid finding from a bounty pool on top of the platform fee, so the total depends on what researchers find. For consultancy work, a US federal regulatory analysis used $10,000 to $30,000 as an estimate for a high-quality professional penetration test. Our penetration testing cost guide breaks down what drives a quote.
What to check behind the headline price
- Scope per unit. “One application” can mean five routes or five hundred. Ask how roles, tenants and API endpoints are counted.
- Platform fees. Synack lists its platform as a separate line item. Confirm whether the quoted price includes access to the findings.
- Expiry. Credits and testing hours that expire at the end of a contract year are a cost if your release schedule slips.
- Retests. How many, how fast, and until when.
- Testing depth. Black box tests are cheaper than gray box tests with accounts. Gray box is the one that finds cross-tenant access, which is what auditors and enterprise customers ask about. Our testing methods guide compares the three.
How to choose a PTaaS provider
Ten questions that separate providers. Each has an answer you can check, so ask for the evidence.
Who tests, and who confirms each finding? Named staff, a crowd, or agents. Ask whether a second person or agent reproduces every finding before it reaches you, because unconfirmed findings cost your engineers triage time.
Can I read a full sample report? Look for scope, methodology, coverage, reproduction steps for each finding, a severity rationale, remediation advice and retest status. A scanner export with a cover page is not a pentest report.
What are the retest terms? Number, turnaround, and how long after the test you can still ask.
How is authorization handled? Signed rules of engagement bound to exact assets and dates, asset ownership checks, and respect for cloud provider policies such as AWS's penetration testing policy.
Do you test authenticated, across roles and tenants? Broken access control is the most common serious finding in web applications. It needs at least two accounts to find.
What is safe by default? Ask what is excluded without your say-so: denial of service, brute force, destructive writes, and testing outside the window.
Where does my data go? Source code, credentials, traffic captures and findings. If AI is involved, ask which model providers see that data, or whether the models run on infrastructure you or the provider control.
Which standards does the report map to? The OWASP Web Security Testing Guide, the OWASP API Security Top 10, and your audit framework. Ask whether an attestation letter is included.
How fast to start, and how fast to a report? Days to start is normal for PTaaS. Ask when the clock starts and what pauses it.
What evidence of skill can you show? For human testers, certifications such as CREST or OSCP. For AI-run testing, results on a public, intentionally vulnerable target that you can check.
For transparency, Helix's answers: AI agents test in isolated sandboxes and a second agent reproduces every finding, with humans overseeing the run. Retests are free and unlimited. Rules of engagement are signed in the portal before any traffic, and testing is detect-only by default. The models run on inference infrastructure we control. Our sample report and our OWASP Juice Shop run show the output on a public target.
Where AI-run testing fits, and its limits
Several PTaaS providers now run part or all of a test with AI agents. Cobalt markets “agentic PTaaS”, and Intruder and Cobalt both list fixed-price AI or autonomous tests. The agents do well on work that is broad and repetitive, and less well where a test depends on context nobody wrote down.
Where agents help
- Mapping every route, role and parameter.
- Trying each known vulnerability class against each of them, the same way every time.
- Retesting a fix in minutes, which is what makes unlimited retests affordable.
- Running at night and on short notice.
Where they fall short
- Business logic that only makes sense with domain knowledge.
- Long attack chains across systems, and judging real-world impact.
- Social engineering and physical testing, which are out of scope for software agents.
- False positives, unless a second pass confirms each finding.
Coverage also varies by model. When we ran the same gray-box test of OWASP Juice Shop with two open-weight models, one confirmed 67 findings and the other 58, and each caught serious bugs the other missed. That is an argument for verification, human oversight and retesting, and for asking any AI-run provider to show results on a target you can check yourself. Our AI penetration testing guide goes further into what the measurements support.
If the application you need tested is itself built on a language model, the attack surface changes. The OWASP Top 10 for LLM applications lists the risks to test for, and our LLM red teaming guide covers how that testing runs.
PTaaS and compliance
Auditors care about the report, not how the test was bought. A PTaaS report works as evidence when it shows the scope, method, findings with evidence, severity and retest results.
- PCI DSS. Requirement 11.4 asks for internal and external penetration tests at least every 12 months and after significant changes, and 11.4.4 asks for exploitable findings to be fixed and retested. See the PCI DSS guide.
- SOC 2. No universal pentest requirement, but a common way to evidence the controls you claim. See the SOC 2 guide.
- ISO 27001. No pentest mandate. Many auditors expect one when your risk assessment relies on it. See the ISO 27001 guide.
The UK NCSC's guidance adds a useful limit: a penetration test “can only validate that your organisation's IT systems are not vulnerable to known issues on the day of the test”. Shipping changes the answer, which is the practical case for testing more often than once a year.
Frequently asked questions
What does PTaaS stand for?
Penetration testing as a service. It describes how a penetration test is bought and delivered: through a platform where you scope the target, authorize the test, receive findings as they are confirmed, and request retests. The testing itself is still a penetration test.
Is PTaaS the same as automated penetration testing?
No. Automated or AI-run testing is one way a PTaaS provider can do the testing. Others use their own staff, a crowd of vetted freelance testers, or a mix of people and automation. Ask each provider who does the testing and who confirms each finding before it reaches you.
How much does PTaaS cost?
Published prices checked on 5 November 2026 start around $3,000 to $3,500 for a single web application test (Helix, Intruder, and Cobalt's autonomous test offer). Researcher-led tests run to five figures: Synack lists $10,283 and $27,120 tiers. Subscriptions such as Astra's run from $2,999 to $9,999 or more per year, and many providers only quote. Compare scope, retest terms, and whether a platform fee or credit expiry applies.
Is PTaaS better than a traditional penetration test?
It is better when you test often, want findings while the test runs, and need retests without a new contract. A traditional consultancy engagement can be the better fit for a one-off, highly bespoke scope such as a red team exercise, a physical test, or work that needs testers on site.
Does a PTaaS report satisfy SOC 2, PCI DSS or ISO 27001?
It can, if the report shows scope, methodology, findings with evidence, severity, and retest results. PCI DSS requirement 11.4 asks for internal and external penetration tests at least every 12 months and after significant changes, with exploitable findings retested after they are fixed. SOC 2 and ISO 27001 do not mandate a pentest, but auditors often accept one as evidence for the controls you claim.
How often should you run a penetration test with PTaaS?
At minimum, as often as your framework requires, which for PCI DSS is every 12 months and after significant changes. Teams that ship weekly often test each major release or each quarter, because a test only shows what was exploitable on the day it ran.
Will AI replace penetration testers?
Not across the board. AI agents already run the repetitive parts of a web or API test well: enumerating routes and roles, trying known vulnerability classes, and retesting fixes. Business logic, novel attack chains, and judgment about real-world impact still benefit from experienced people, which is why credible AI-run services keep humans overseeing the work.
Is penetration testing legal?
Yes, when the owner of the systems has given written permission for that scope and time window. Testing without authorization can be a crime under laws such as the US Computer Fraud and Abuse Act and the UK Computer Misuse Act 1990. A PTaaS platform should make you sign rules of engagement before any traffic is sent.
Sources
- NIST CSRC glossary: penetration testing (from SP 800-115)
- UK NCSC: penetration testing guidance
- PCI Security Standards Council document library (PCI DSS v4.0.1)
- OWASP Web Security Testing Guide
- Bugcrowd: bug bounty glossary
- HackerOne: what is penetration testing as a service
- Synack: what is PTaaS
- Vendor pricing pages, checked 5 November 2026: Cobalt, Synack, Astra, BreachLock, Intruder
Prices change. Each figure was read from the vendor's public page on the date above; confirm the current terms with the vendor before you compare quotes.