Pentest as a service (PTaaS): what it is, what it costs, who needs it

Pentest as a service (PTaaS) is human-led penetration testing delivered through a platform: you scope the test online, findings appear while the testers work, and retests are run from the same dashboard. Published prices start near $5,000 for one small web app. It suits a company that needs a tester's report several times a year. Between two tests, automated attacks on every deploy cover what changed.

The team behind Keybrum's attacksUpdated 10 min read

In short

  • PTaaS is a delivery model. Bugcrowd defines it as human-led penetration testing combined with a cloud platform, and says it does not mean replacing pentesters with a scanner.
  • It is sold as tests or as yearly credits. Bugcrowd lists $5,000 to $15,000 per web app test on a 12-month contract, and Cobalt sells credits worth 8 hours of testing each.
  • What you gain over a one-off test is speed and retests: Bugcrowd says a test launches within three business days, and Cobalt includes retesting for the length of the contract.
  • "Continuous pentesting" has two meanings: human tests run several times a year on a subscription, or software that attacks every build. Ask which one a vendor sells.
  • Keybrum is the automated end of this range. It attacks every deploy and proves each finding, but it has no human tester and does not issue a pentest firm's attestation letter.
On this page

What is pentest as a service?

Pentest as a service, or PTaaS, is a penetration test bought and run through an online platform instead of a consulting contract. A penetration test is an attack on your app that you asked for, by people who then report what they could reach. With PTaaS the people are still there. What changes is everything around them: you describe the scope in a web form, the vendor assigns testers from its pool, findings show up in a dashboard while the test is running, and you ask for a retest with a button.

Bugcrowd, one of the vendors, defines PTaaS as "a delivery model that combines human-led penetration testing with a cloud platform", and adds that it does not mean replacing pentesters with a scanner. Cobalt, which calls itself a pioneer of the model, describes it as "a platform-delivered model for offensive security testing". Both definitions say the same thing: the testing is the same craft, the delivery is software.

That matters when you compare offers, because the label is used loosely. Some products sold under the name are a scanner with a dashboard. The question that sorts them is simple: who attacks the app, a named person or a program?

PTaaS, a one-off pentest and automated attacks compared

Three kinds of offers answer the same search. They differ in who does the work, how fast you get results and what the report is worth to a third party.

One-off manual pentestPTaaSAutomated attacks on every deploy
Who attacksNamed testers from a consultancy, for a fixed number of daysTesters from the platform's pool, for the hours you boughtSoftware, with no person deciding each request
How you buyA quote per engagement, after a scoping callPer-test units or yearly credits, scoped in the platformPer run or a monthly plan
Time to startOften booked weeks aheadBugcrowd: within three business days. Cobalt: human-led tests launch within 24 hoursMinutes
When you see findingsIn the final reportDuring the test, in the dashboardDuring the run
RetestsOften one round in a fixed window, sometimes billedUsually included for the contractAfter each fix and on every deploy
Business logic, chained attacksYes, the main strength of a personYes, as deep as the hours boughtOnly the patterns the tool has attacks for
Report an auditor or a buyer acceptsThe usual choiceUsually accepted, check the format firstAsk the auditor before you rely on it
What goes staleEverything shipped after the testEverything shipped between two testsNothing shipped, but only within the tool's library

Bugcrowd's own guide is fair about the limit of the model: an organization with one stable, narrowly scoped system may still be well served by a traditional project. PTaaS pays off as the number of apps, the release rhythm and the number of tests a year go up.

How PTaaS is priced

Vendors use two units. Some sell tests: a fixed price for one app of a given size. Others sell credits for a year and debit them per engagement. Here is what vendors print on their own pages in November 2026. The pentest cost guide has the full comparison, with manual day rates.

VendorUnitPublished priceRetests
Bugcrowd Standard Pen TestOne web app, by complexity, on a 12-month contract$5,000 small, $8,000 medium, $15,000 large12 months of retesting with one report update
CobaltCredits sold in yearly packages; one credit is 8 hours of testingOn quote. Credits do not roll over to the next contractUnlimited on-demand retesting during the contract
Astra Pentest ExpertOne target, manual test by its testers plus its scanner$5,999Two rescans within 30 days
SynackYearly programSynack's guide puts annual PTaaS at $20,000 to $100,000 or moreIncluded, per its guide
Prices before tax, read on each vendor's site on 13 November 2026. Bugcrowd's are its AWS Marketplace list prices. Vendors change prices often: check the source before you budget.

Synack's pricing guide states the trade plainly: a yearly program carries a higher sticker price than a single engagement, and it covers retests and repeated validation that a one-off test bills separately. Its example of the alternative is a $10,000 engagement that charges $3,000 per retest, which runs over budget when you fix in several batches.

Read the unit before the price. A credit model is flexible if you have several apps to test through the year. It is expensive if you buy a package for one app and the unused credits expire.

What continuous pentesting means, and what it does not

Continuous pentesting is the phrase that confuses most buyers, because two different products carry it.

  • Human tests on a subscription. Testers come back several times a year, or work through a list of assets across the year. Each test is still a point in time. Between two tests, what you ship is not looked at by anyone.
  • Software on every build. A tool attacks the app on a schedule or at each deploy. It never gets tired, and it only knows the attacks in its library. Several vendors now sell this next to their human tests: Cobalt lists an Autonomous Pentest at $3,500 per test as a limited-time offer through 31 December 2026, Astra lists Pentest Auto at $2,999 per pentest, and Intruder lists an AI pentest that reads your repository at $3,500 per test.

Neither is a trick. They answer different questions. The first asks what a skilled person finds in your app this quarter. The second asks whether yesterday's deploy opened something a known attack can reach. The automated penetration testing guide covers what software can prove and where it stops.

Who needs which

Your situationWhat fitsWhy
Building or just launched, nobody asks for a reportSelf-checks, then automated attacksThe common leaks in generated apps are open tables, keys in the bundle and missing ownership checks. Software finds those the day they ship
One app, one enterprise customer asking for a pentest reportA one-off manual test or a single PTaaS testYou need one dated report from a named firm, with a retest
Several apps or a release every week, and customers or auditors who ask every yearPTaaSScoping once, tests through the year and retests in one place cost less effort than four contracts
Card data in scope or regulated dataA qualified tester on the schedule your standard setsSee PCI DSS penetration testing
Any of the above, between two human testsAutomated attacks on every deployA human test describes one build. Your app changes after it

The pentest guide for startups goes through the timing in detail. The short version: buy a human test when someone needs the report, and keep something attacking the app in between.

Checks to run before you pay for tester hours

On a credit model, an hour a tester spends reporting an open table is an hour not spent on your refund flow. Close what anyone can find first. If your app runs on Supabase, this read-only query in the SQL editor lists the tables that answer to anyone holding your public key:

sql
-- Tables in the public schema with row level security off
select tablename
from pg_tables
where schemaname = 'public' and not rowsecurity;
Run it in your own project. It only reads the catalog. Every table it returns needs RLS and policies.
  1. 1

    Close open tables and rules

    Enable row level security on every table the query returns, then write policies. The Supabase RLS guide has them ready to paste. On Firebase, check that no rule is still in test mode.

  2. 2

    Look for server keys in your JavaScript

    Paste your built bundle into the API key leak checker. A Stripe secret key or a Supabase service role key in the browser is a finding you do not need a tester for.

  3. 3

    Run the two-account test

    Create user A and user B on your own app. As B, ask for a record A created, by changing the id in the address or in the API call. If B gets it, that is an IDOR, the most common serious finding in these apps.

  4. 4

    Write the scope and create the accounts

    List the domains, the API routes and the roles. Create two test accounts per role. Testers on any platform go deeper when they do not spend the first hours getting in.

Questions to ask a PTaaS vendor

  • Who tests: named people, a pool, or software? Which findings does a person confirm?
  • What is the unit: a test, a credit, an asset? What does one credit buy in hours, and do unused credits expire?
  • How many retests are included, within what window, and what does one more cost?
  • How soon can a test start after I submit the scope?
  • Will the report and the attestation letter satisfy the customer or auditor who asked? Ask for a sample.
  • What happens between two tests: is anything watching what I deploy?
  • What is left out of scope: rate limits, third-party APIs, flows that need a real card?

The sample pentest report shows what a buyer's security team expects to read, and the security questionnaire guide covers what to answer while you have no report yet.

Where Keybrum sits: the automated end, said plainly

Keybrum is an automated attack tool for apps built with AI tools such as Lovable, Bolt, Cursor and Claude Code. It is the third column of the table above. It is not a human-led PTaaS: no tester reads your app, and it does not issue the attestation letter of a pentest firm.

What it does: it attacks your live app as a stranger and as two accounts it signs up. Its library covers Supabase tables, database functions, storage and realtime channels, Firebase rules, secret keys in the JavaScript bundle (each checked with a read-only call to the provider), one account reading or changing another account's records, AI endpoints that answer without an account, unsigned Stripe webhooks, exposed .env files and known Next.js and React CVEs. Every finding carries the request that worked. Your coding agent reads the fix through Keybrum's MCP server, and the finding closes only when the same attack fails.

  • One-time report, €49: one app attacked once, proof of every finding, the fix for your coding agent, retests of those findings for 7 days, and the PDF report.
  • Plans: Maker at €29 a month for 5 apps, Team at €99 a month for 25 apps with your brand on reports and trust pages. Plans attack every deploy, attack again every week or month, and keep retests open. Prices exclude VAT.
  • Only your own apps: Keybrum attacks an app after its owner proved ownership.

What it does not do: judge business logic the way a person does (a refund issued twice, a coupon stacked, a checkout step skipped), test anything outside its library, run social engineering, or produce the signed report an auditor may require. If a customer contract asks for a third-party penetration test by a named firm, buy that test. Run Keybrum before it, so the tester's hours go to what software cannot find, and after it, so the next deploy is attacked too.

Questions

What does PTaaS stand for?

PTaaS stands for penetration testing as a service, also written pentest as a service. It means a penetration test ordered, followed and retested through an online platform. The testing itself is still done by people in the offers that use the term strictly.

How much does pentest as a service cost?

Published prices start near $5,000 for one small web app: Bugcrowd lists $5,000, $8,000 and $15,000 by complexity on a 12-month contract, and Astra lists its expert plan at $5,999. Synack's guide puts yearly PTaaS programs at $20,000 to $100,000 or more. Cobalt sells yearly credit packages on quote.

Is PTaaS the same as a vulnerability scan?

No. A vulnerability scan matches versions and patterns and reports what might be wrong. PTaaS, in the strict sense, has human testers who exploit what they find and report what worked. Some products use the PTaaS label for a scanner with a dashboard, so ask who does the testing.

What is continuous pentesting?

It means one of two things. Either human testers return several times a year under a subscription, or software attacks the app on every build or on a schedule. The first goes deeper on each visit. The second sees every change, within the attacks it knows.

Does a PTaaS report work for SOC 2 or a customer questionnaire?

Usually, because the test is done by people from a named firm, but formats differ. Ask the vendor for a sample report and an attestation letter, and show them to your auditor or your customer's security team before you sign.

Is Keybrum a PTaaS?

No. Keybrum is automated: software attacks your live app on every deploy and proves each finding with the request that worked. It has no human tester and no attestation letter from a pentest firm. It is what runs between two human tests, from €49 for a one-time report.

Sources

  1. 1.What is penetration testing as a service?, Bugcrowd
  2. 2.Bugcrowd Platform: pricing information (Standard Pen Test), AWS Marketplace
  3. 3.Pentest as a service, Cobalt
  4. 4.Pricing and credits, Cobalt
  5. 5.Pricing, Astra Security
  6. 6.How much does a pentest cost? (2026 pricing guide), Synack
  7. 7.AI pentesting: pricing, Intruder

Your app is public. So are its mistakes.

Check what any attacker can already see in it, free, in ten seconds. The full attack, with proof and fixes: €49 once, or €29 a month on a plan.

or continue with Google or GitHub