Skip to content

Continuous penetration testing, every week

A yearly pentest describes software that no longer exists. Grapefruit maintenance plans re-test your app, APIs and AI features every week or every month. Each run is dated, kept, and shows what is new and what got fixed. For teams that deploy faster than they audit.

What is continuous penetration testing?

Continuous penetration testing is a pentest that repeats on a schedule instead of once a year. With a Grapefruit maintenance plan, your verified domains are attacked again every week or every month. Each run is dated and kept in your portal, and each one shows what is new and what got fixed since the last.

It is the same real exploitation as a full test, covering your web app, your REST and GraphQL APIs, your LLM features and your AI agents. The difference is rhythm. Your security evidence keeps pace with your release notes.

Why is a yearly pentest out of date so quickly?

Because your product changes every week and the report does not. A team that deploys daily ships hundreds of changes between annual tests. By the time the PDF arrives, the code it describes has been rewritten, new endpoints have appeared, and the AI feature you launched last quarter was never in scope.

The calendar makes it worse. According to Stingrai, traditional engagements commonly take 2 to 4 weeks from accepted quote to start, 5 to 10 business days of testing, and the report 1 to 2 weeks after that. That is a snapshot delivered a month or two late, then left to age for a year.

Fixes drift too. Cobalt's 2025 State of Pentesting data, as reported by CSO Online, found that 48% of pentest findings get fixed, and 69% of serious ones. With a weekly run, an open finding shows up again in every report until it closes.

How does the weekly loop work?

Each week Grapefruit attacks your verified domains again, compares the results with the previous run, and updates your portal. New findings are flagged as new. Findings that no longer reproduce are marked fixed. Everything else carries forward with its history. You open one dated run and see exactly what changed.

  • Test: your app, APIs and AI features are attacked with the test accounts you supplied.
  • Compare: results are matched against the last run, finding by finding.
  • Report: the portal shows what is new, what got fixed and what is still open, ranked critical, high, medium or low.
  • Replay: every new finding carries the exact request and response, so your engineer can run it and watch it work.
  • Fix: you ship the patch. The next scheduled run tests it and marks the finding fixed.

What changes does continuous testing catch?

It catches the things that happen between audits: a new endpoint shipped without an ownership check, a fix that quietly came undone in a refactor, an AI feature that went live on a Tuesday. Each is found on the next run, days after it shipped, with a request that proves it.

  • A new endpoint. Last sprint added an export route. It checks that the caller is logged in and forgets to check whose data is requested. The next run finds it and flags it as new.
  • A regressed fix. A finding closed in March. In June a refactor moves the permission check and the old attack works again. The run reports it as reopened, with the original request still attached.
  • A new AI feature. You add an assistant that can look up orders. The next run attacks it for prompt injection and for tool calls that reach other customers' records, together with the API behind it.
  • A changed role. A new 'billing viewer' role ships with one permission too many. The run exercises it against every action the admin role has.

How does an annual pentest compare with continuous re-testing?

An annual pentest gives you one report about one week of your product. Continuous re-testing gives you a dated record of every week. The annual test still has its place as a deep baseline, and Grapefruit sells a full penetration test at a fixed quote for exactly that. The plan keeps the baseline true.

Annual pentestContinuous re-testing
How oftenOnce a yearEvery week or every month
What it describesThe app as it stood during the test windowThe app as it stands this week
Time to resultsReport 1 to 2 weeks after testing ends (Stingrai)Each run posts to the portal, first results within hours of kickoff
New featuresTested at the next annual engagementTested on the next scheduled run
Fix verificationA re-test you schedule, or next year's testRe-tested on the next run and marked fixed
RegressionsFound a year later, if at allFlagged as reopened on the next run
Evidence trailOne dated reportA dated run for every week or month of the period
BuyingQuote, scheduling, statement of workPublic price, card checkout, start today

How does the dated history serve a SOC 2 Type II window?

A Type II audit looks at a period, and a dated run history covers a period. According to Baker Tilly, SOC 2 does not formally require a penetration test, but auditors commonly expect one, and for Type II the testing has to be evidenced inside the audit period. Weekly runs put evidence in every week of it.

When the auditor asks what you tested and when, you open the portal. Each run has a date, a list of findings by severity, the request and response behind each one, and the run in which it was marked fixed. That is detection and remediation on the record, with the timeline already built.

Weekly or monthly: which plan fits?

Pick the cadence that matches how often you ship. Teams that deploy several times a week, or that are adding AI features quickly, choose weekly so a new flaw lives for days. Teams with a slower release train choose monthly and still get twelve dated runs a year instead of one.

Both plans cover the same ground with the same replayable evidence, and both have public prices. See pricing to compare them.

How do you start?

Check out with a card, verify your domain, and testing is running. No demo call and no scheduling. The whole setup takes a few minutes, and the first run's findings land in your portal within hours of kickoff. From then on the schedule runs itself.

  • Choose weekly or monthly and pay by card.
  • Add your hostname and publish one DNS TXT record to verify it.
  • Sign the testing authorization online and add your test accounts.
  • Read the first run the same day. The next one arrives on schedule.

Questions

Is continuous penetration testing the same as vulnerability scanning?
No. A scheduled scan lists possible issues from the outside. Each Grapefruit run is autonomous testing with real exploitation, signed in with your test accounts, and a finding is reported with the captured request and response that proves it. Your engineers can replay every one and see the attack work.
How often are we re-tested?
Every week or every month, depending on the maintenance plan you choose. Each run covers your verified domains, is dated and kept in the portal, and shows what is new and what got fixed since the previous run. Pick the cadence that matches how often you ship. See pricing for the plans.
What happens when we fix a finding?
The next scheduled run tests it again. If the original attack no longer works, the finding is marked fixed and the run records the date. If a later change brings the flaw back, a future run reports it again with the same evidence, so regressions surface in days.
Do we still need a full penetration test?
Many teams start with one. A full penetration test at a fixed quote gives you a deep baseline and a report to share with customers. The maintenance plan then keeps that baseline current with weekly or monthly runs, so the next time someone asks for your latest test, it is from this week.
Will auditors accept the run history as SOC 2 evidence?
Each run is dated and includes findings by severity, the request and response evidence, and the fix status, with a downloadable report. Baker Tilly notes that auditors commonly expect a penetration test and that for Type II the testing has to be evidenced inside the audit period. A weekly record does that for every week.
Does continuous testing cover our AI features?
Yes. LLM features such as chat and retrieval, and AI agents with tools, are tested in every run together with the web app and APIs. When you ship a new assistant or give an agent a new tool, the next scheduled run attacks it and reports the results in the same portal.
How much does it cost?
Maintenance plans have public prices and card checkout, with no demo call. See pricing for the weekly and monthly figures. For context, Astra's cost guide puts a single traditional web application pentest at $5,000 to $30,000, more for complex multi-role apps.

Standards and sources