What is web application penetration testing?
Web application penetration testing is a controlled attack on your live app that proves which weaknesses can really be exploited. Grapefruit signs in with test accounts you supply, attacks every role and workflow, and hands your engineers ranked findings, each with the exact request and response that proves it, within hours of kickoff.
The point is evidence. A scanner tells you a page might be vulnerable. A penetration test shows the invoice that leaked, the account that was taken over, the admin action a normal user completed. You get the second kind, ranked critical, high, medium or low, in an online portal and a downloadable report.
What does a web application pentest cover?
Coverage follows the OWASP Web Security Testing Guide (WSTG) v4.2, the checklist auditors and enterprise buyers recognize. Grapefruit attacks authentication, session management, authorization between roles and tenants, input handling, business logic, file handling and client side code, then reports what was actually exploited, with the traffic that shows it.
Multi-tenant products live or die on authorization. Supply accounts in more than one role and more than one tenant and every boundary between them gets attacked: user to admin, customer to customer, free plan to paid feature.
| Area | What we attack | Example finding |
|---|---|---|
| Authentication | Login, signup, password reset, multi-factor steps, single sign-on flows | A reset link that still works after the password was changed |
| Session management | Cookies, tokens, logout, session expiry, concurrent sessions | A session token that stays valid after logout |
| Authorization between roles and tenants | Every action and record, requested as the wrong role or the wrong customer | A viewer account exports the full user list of another tenant |
| Input validation and injection | Forms, search, filters, headers and any field that reaches a database, template or shell | A search filter that returns rows straight from another table |
| Business logic | Checkout, coupons, approvals, invites, plan limits, multi-step workflows run out of order | An approval step skipped by calling the final step directly |
| File handling | Uploads, downloads, imports, exports, generated documents | A download link that serves any customer's file when the name is changed |
| Client side | Scripts, stored content, redirects, cross-origin settings, browser storage | A comment field that runs script in an admin's browser |
How is this different from a traditional pentest?
A traditional engagement runs on a consultant's calendar. Grapefruit runs on yours. According to Stingrai, traditional pentests commonly take 2 to 4 weeks from accepted quote to start, 5 to 10 business days of testing, and the report lands 1 to 2 weeks after that. Grapefruit starts the day you verify your domain.
The cost picture changes too. Astra's penetration testing cost guide puts traditional web application pentests at $5,000 to $30,000, more for complex multi-role apps. Grapefruit gives you a fixed quote for a full test and public prices for maintenance plans. See pricing for both.
| The old way | Grapefruit | |
|---|---|---|
| Lead time | 2 to 4 weeks from accepted quote to start (Stingrai) | Starts the day you verify your domain |
| Report wait | 1 to 2 weeks after testing ends (Stingrai) | Results within hours of kickoff |
| Cost | Commonly $5,000 to $30,000, more for complex multi-role apps (Astra) | Fixed quote for a full test, public prices for maintenance plans |
| Proof | A written description your engineer has to reproduce | Every finding replayable from the captured request and response |
| AI features | Usually a separate engagement from a separate vendor | Included in the same test |
What does a Grapefruit finding look like?
Every finding is a replayable proof. You get the exact request that was sent and the exact response your app returned. Your engineer pastes the request, runs it, and watches the attack work. After the fix ships, they run it again and watch it fail. No debate about whether the issue is real.
That matters more than it sounds. In a Pentest-Tools survey of 158 respondents in June 2026, 87.8% said they had hit AI pentest findings that needed significant manual validation. Replayable evidence removes that step. The finding validates itself in under a minute.
- Severity: critical, high, medium or low, so the queue sorts itself.
- Where: the endpoint, the parameter and the role that was used.
- Proof: the captured request and the response that came back.
- Impact: what an attacker gets, in plain language a product manager can read.
- Status: open or fixed, updated when the fix is re-tested.
Why test the app, the API and the AI features together?
Real attacks do not stay in one layer, so the test should not either. Grapefruit attacks your web app, your REST and GraphQL APIs, your LLM features and your AI agents in one run and reports them in one place. A chain that crosses all three shows up as one finding, start to finish.
Picture a support ticket. An attacker types instructions into the ticket body. Your AI assistant reads the ticket for a staff member, follows the hidden instructions, and calls an internal API with the staff member's permissions. A web-only test sees a harmless text field. An AI-only test never sees the API. One combined test sees the breach.
How fast can you start?
Today. Setup takes a few minutes: add your hostname, publish one DNS TXT record to verify the domain, and sign the testing authorization online. Add test accounts for the roles you want covered. Testing starts, and the first findings reach your portal within hours of kickoff.
- Add the hostname of the app you want tested.
- Publish one DNS TXT record. That verifies the domain is yours.
- Sign the testing authorization online.
- Add test accounts for each role, and for two tenants if your product is multi-tenant.
- Watch findings arrive in the portal, ranked and ready to replay.
What happens after you fix a finding?
The fix gets re-tested and the finding is marked fixed when the same attack stops working. Your engineers can replay the original request themselves the moment the patch deploys, so nobody waits on a vendor to learn whether the fix held.
Most teams then keep the app on a maintenance plan. Your verified domains are re-tested every week or every month, each run is dated and kept, and each one shows what is new and what got fixed since the last. The app you shipped on Friday gets attacked before your customers find the gap.
Questions
- How long does a web application penetration test take?
- Results land within hours of kickoff. There is no scheduling queue: you add your hostname, verify it with one DNS TXT record, sign the authorization online and testing starts the same day. Traditional engagements commonly take 2 to 4 weeks just to begin, according to Stingrai, plus another 1 to 2 weeks for the report after testing ends.
- Do you test behind the login?
- Yes. Authenticated testing is the core of the service. You supply test accounts, ideally one per role and accounts in two separate tenants, and Grapefruit attacks the app as each of them. That is how authorization flaws are found: one customer reading another customer's data, or a basic user completing an admin action.
- How much does web application penetration testing cost?
- A full penetration test is a fixed quote, so you know the number before you sign. Maintenance plans with weekly or monthly re-tests have public prices and card checkout. See pricing for current figures. For comparison, Astra's cost guide puts traditional web application pentests at $5,000 to $30,000, more for complex multi-role apps.
- What do I get at the end?
- An online portal with every finding ranked critical, high, medium or low, each with the captured request and response that proves it, plus a downloadable report you can hand to customers, auditors or your board. Fixed findings are re-tested and marked fixed, so the report reflects where the app stands now.
- Is this a vulnerability scan?
- No. A scan lists things that might be wrong. Grapefruit performs autonomous testing with real exploitation, which means a finding is only reported with the request that worked and the response that came back. Your engineers can replay each one and see the result for themselves in under a minute.
- Our app has a chatbot and an AI agent. Is that a separate test?
- No, it is the same test. LLM features such as chat and retrieval, and AI agents with tools, are attacked alongside the web app and its APIs and reported in the same portal. Attacks that start in a text field and end in an agent's tool call are found as one chain.