Short answer
Choose the kind of provider before you choose the company. Consultancies such as Bishop Fox and NCC Group suit large, unusual or high-stakes scopes. PTaaS platforms such as Cobalt, Synack and HackerOne suit teams that test often and want findings in a portal. AI-led testers such as Helix Cyber, XBOW and Intruder report within hours or a day at published prices. Tools such as NodeZero and Pentera are software your own team runs.
Closest fit by need
- Red teaming, hardware, mainframe or OT, with senior consultantsBishop Fox or NCC Group
- Human testers on demand, many assets a yearCobalt, HackerOne or Bugcrowd
- Government work with FedRAMP authorizationSynack
- Manual testing with prices on the websiteSoftware Secured
- One app, API, network or cloud account tested this week, every finding reproducedHelix Cyber
- Your own team testing internal networks and Active Directory every weekHorizon3.ai NodeZero
On this page · 11 sections
Helix CyberFrom $3,000See Helix Cyber
Four kinds of penetration testing provider
The same search returns companies that sell very different products. The differences that matter most are who does the testing, how you pay, and what you get back.
Consultancy
- Who tests
- Employed consultants
- How you buy
- Quoted per engagement
- What you get
- A report weeks after booking; strongest on novel, complex and physical scopes
PTaaS platform
- Who tests
- Vetted testers, often freelance, through a platform
- How you buy
- Credits, subscriptions or packages
- What you get
- Findings in a portal within days to two weeks; strongest on breadth across many assets
AI-led tester
- Who tests
- AI agents, with or without human review
- How you buy
- Per test, often at a published price
- What you get
- A report in hours to about a day; strongest on speed, price and repeat testing
Tool you run
- Who tests
- Software, operated by your own team
- How you buy
- Annual licence, often per asset
- What you get
- Results as often as you run it; strongest on internal networks between tests
The lines blur. Bishop Fox runs a continuous platform, NetSPI sells PTaaS with employed testers, and most PTaaS platforms now add AI agents. Group a provider by how it does most of its testing.
What the providers that publish prices charge
Seven of the 17 publish a price for a pentest. These are starting prices for the smallest scope each one lists, in US dollars, from their own pages on November 5, 2026. For what moves a quote up or down, see our penetration testing cost guide.
| Provider | Product | Starting price |
|---|---|---|
| Helix CyberOurs | Web app or API: black, gray or white box | $3,000 · $4,000 · from $7,000/test |
| Intruder | AI web app pentest | $3,500/test ($4,000 standalone) |
| Cobalt | Autonomous Pentest, web apps | $3,500/test (promo to Dec 31, 2026) |
| Synack | Standard Sara (AI) pentest · Standard pentest | from $5,085/test · from $10,283/test |
| Astra Security | Pentest Expert (manual) | $5,999/yr |
| Software Secured | Web and API pentest (manual) | from $10,800/test |
| Horizon3.ai | NodeZero Core · Flex · WebApp Flex | $50/asset/yr · $15/asset/test · $10,000/asset/test |
All 17 penetration testing companies at a glance
| Provider | Who tests | Starting price | Retests |
|---|---|---|---|
| Consultancies · 5 | |||
| Bishop Fox | Employed consultants plus a managed platform | Not published | Not stated |
| NCC Group | In-house consultants | Not published | Not stated |
| Rapid7 | In-house consultants | Not published | Not stated |
| TrustedSec | In-house consultants | Not published | One retest after fixes |
| Software Secured | Employed, certified human testers | From $10,800/test (web and API) | 3 rounds over 12 months; unlimited on PTaaS |
| PTaaS platforms · 7 | |||
| Cobalt | Freelance Cobalt Core pentesters, AI-assisted | Quoted credits; Autonomous Pentest $3,500/test (promo to Dec 31, 2026) | Unlimited during the contract term |
| Synack | Vetted contract researchers; AI (Sara) with human validation | From $5,085/test (Sara); $10,283/test (Standard) | Patch verification included; count not stated |
| NetSPI | Employed testers (350+) | Not published | Included; rounds not stated |
| HackerOne | Vetted community pentesters; optional AI agents | Not published | Included; count not stated |
| Bugcrowd | Curated crowd testers | Not published | 12 months (Standard) |
| BreachLock | In-house certified testers plus AI | Not published | 1 to 2 manual retests, or unlimited (pages differ) |
| Astra Security | AI agents (Auto) or certified testers (Expert) | From $199/mo (Auto); $5,999/yr (Expert) | 1 to 4 re-scans by plan |
| AI-led testers · 3 | |||
| Helix CyberOurs | AI agents, overseen by people; a second agent reproduces each finding | $3,000/test (web app or API, black box) | Free and unlimited |
| XBOW | Autonomous AI agents | Not published | Free and unlimited |
| Intruder | Scanners plus AI agents; no human tester | Free plan; AI pentest $3,500/test | Unlimited (AI pentest) |
| Tools you run · 2 | |||
| Horizon3.ai NodeZero | Autonomous software you run | $50/asset/yr (Core); $15/asset/test (Flex) | Unlimited on annual plans |
| Pentera | Automated platform you run (or managed) | Not published | Re-run on demand |
Consultancies
Consultancies employ their testers and scope each engagement with you. They are the natural choice for red teaming, social engineering, hardware and anything without a template. None of the four large firms here publishes prices; Software Secured does.
Bishop Fox
Consultancy with a continuous platform- Best for
- Large enterprises that want senior consultants across apps, cloud, hardware and AI, plus continuous external testing.
- Pricing
- Not published.
Strengths
- Testing across applications, networks, products, cloud, hardware and AI systems, plus red teaming and social engineering.
- Cosmos, its continuous testing platform, is run by Bishop Fox and passes findings through human validation.
- Authors of widely used offensive tools such as Sliver and CloudFox.
Check before you buy
- Retest terms are not stated on the pages we read.
NCC Group
Global consultancy- Best for
- Multinational and regulated organisations that want one provider for applications, cloud, hardware, cryptography and AI.
- Pricing
- Not published.
Strengths
- One of the widest catalogues: network, web and mobile apps, cloud, hardware, blockchain, cryptography and AI testing, plus red, purple and black teaming.
- Fully automated, semi-automated and manual options at different budgets, and a continuous testing service.
- Findings delivered through its Cyber Services Portal.
Check before you buy
- Retest terms are not stated.
- Accreditations appear as logos only; ask for the certificates your contract needs.
Rapid7
Consultancy within a security platform- Best for
- Rapid7 customers who want a point-in-time network, web, IoT, wireless or social engineering test from the same vendor.
- Pricing
- Not published.
Strengths
- Consultants spend up to 20% of bench time on attacker research, and contribute to the Metasploit Project.
- A clear list of point-in-time tests: network, web app, IoT, social engineering, red team and wireless.
- Continuous red-team operations are available as a separate service.
Check before you buy
- Mobile, API, cloud and AI testing are not listed on the pentest page.
- Retest terms are not stated.
TrustedSec
Consultancy- Best for
- Organisations that want a consultant-led engagement scoped on a call, including staff awareness testing.
- Pricing
- Not published.
Strengths
- A retest after remediation is part of the engagement model.
- 7,400 custom engagements completed and a 92% Net Promoter Score, per TrustedSec.
- Supports PCI DSS, HIPAA and SOC 2 testing requirements.
Check before you buy
- The pentest page lists outcomes, not test types; ask for the scope list.
Software Secured
Manual pentest firm with PTaaS- Best for
- SaaS teams that want human-led manual testing with prices published before the first call.
- Pricing
- Published starting prices; a fixed price follows a scoping call.
- Web and API
- from $10,800/test
- Mobile app
- from $5,400/test
- External network
- from $5,400/test
- Internal network
- from $7,700/test
- Cloud review
- from $6,200/test
- AI pentest
- from $10,800/test
- PTaaS subscription
- from $21,400
Strengths
- The most complete published price list of any manual testing firm here.
- Certified testers (OSCP, OSWE, GWAPT and cloud certifications), “never contractors”.
- Retesting is built in: three rounds over 12 months for most services, unlimited on PTaaS.
- Covers web, API, mobile, network, cloud, AI, IoT and hardware, plus red teaming and threat modelling.
Check before you buy
- Prices are starting points; a fixed price is confirmed after a scoping call.
- External network tests include one retest round, not three.
PTaaS platforms
Penetration testing as a service moves ordering, findings and retests into a web platform. Most draw testers from a vetted community, and most now add AI agents for reconnaissance or validation. NetSPI and Synack completed a merger in October 2026 and still sell separately.
Cobalt
PTaaS platform- Best for
- Teams that test many assets a year and want human pentesters on demand across web, mobile, network, cloud and AI apps.
- Pricing
- Credits on Standard, Premium and Enterprise contracts, quoted on request. One credit is eight hours of testing. Autonomous Pentest for web apps is $3,500 per test as a limited-time offer for tests completed by December 31, 2026.
Strengths
- A network of 500+ vetted freelance pentesters (the Cobalt Core), with a five-stage vetting process and background checks.
- Human-led tests launch within 24 hours from an in-platform scoping wizard, according to Cobalt's PTaaS page.
- Unlimited on-demand retesting during the contract term.
- 50+ integrations, including Jira, GitHub and GitLab.
- Covers web, API, mobile, desktop, internal and external network, cloud and AI or LLM applications, plus red teaming and code review.
Check before you buy
- Cobalt's pricing FAQ says credits do not roll over into the next contract.
- Start times differ by page: within 24 hours on the PTaaS page, one to three business days by tier on the pricing page.
- The $3,500 Autonomous Pentest price is a promotion with an end date.
Synack
PTaaS platform · now part of NetSPI- Best for
- Government and regulated buyers who need a FedRAMP-authorized platform and a large vetted researcher pool.
- Pricing
- Published starting prices. The Synack Platform is a separate line item; a Basic Platform is free.
- Standard Sara (AI) pentest
- from $5,085/test
- Standard pentest
- from $10,283/test
- Synack14 pentest
- from $27,120/test
Strengths
- The Synack Red Team: 1,500+ vetted researchers who pass a five-step vetting process.
- FedRAMP Moderate Authorized, and available through Carahsoft and GSA Advantage.
- One of the few providers here that publishes starting prices.
- Sara, its AI pentester, has every finding validated by a human researcher.
Check before you buy
- Sara currently tests external web and host assets only, and does not yet support MFA or one-time passwords.
- NetSPI and Synack closed their merger in October 2026 and still sell separately; NetSPI says existing engagements are unchanged. Ask how contracts and roadmaps will change.
- Researchers are independent contractors, not employees.
NetSPI
PTaaS and services · merged with Synack- Best for
- Enterprises with unusual surfaces (mainframe, OT, hardware, thick clients) that want one provider and employed testers.
- Pricing
- Not published.
Strengths
- 50+ pentesting services, from web and cloud to mainframe, ATM, automotive, medical and OT hardware.
- 350+ in-house security experts, described as “employed, not outsourced”.
- The NetSPI Platform adds attack surface discovery and breach and attack simulation packs.
- Testing since 2001.
Check before you buy
- Remediation testing is listed, without a stated number of rounds.
- NetSPI and Synack closed their merger in October 2026 and still sell separately; NetSPI says existing engagements are unchanged. Ask how contracts and roadmaps will change.
HackerOne
PTaaS platform- Best for
- Teams already running a HackerOne bug bounty or VDP who want pentests on the same platform.
- Pricing
- Not published; quoted by sales.
Strengths
- Two products: H1 Agentic Pentest pairs AI agents with human pentesters, and H1 Pentest can run fully human-led with no agents.
- Human pentesters review every agent finding, per HackerOne.
- Testers matched to your asset type and stack from a vetted community.
- Covers web, API, cloud, AI and LLM, mobile, network and desktop.
- Findings arrive live, with GitHub, Jira, Slack and ServiceNow integrations.
Check before you buy
- Start times are not published.
- Retesting is included, without a stated count or window.
- The agentic product currently covers web apps and APIs only.
Bugcrowd
PTaaS platform- Best for
- Teams that want a curated tester team, a long retest window, or a pay-for-impact model alongside flat-rate tests.
- Pricing
- Not published on the pages we read. Flat-rate Standard, Plus and Max packages, plus a pay-for-impact model.
Strengths
- Launch in under 72 hours; the Standard package says within three business days.
- Standard includes 12 months of retesting, with one report update.
- CrowdMatch assembles a vetted team for your stack and can rotate testers.
- CREST accreditation for pen testing.
Check before you buy
- Ask for the subscription datasheet; it is linked from the PTaaS page but we could not read it.
- The AI Pen Test and PTaaS pages describe different package tiers.
BreachLock
PTaaS platform- Best for
- Buyers who want in-house certified testers plus an AI engine that asks before it moves laterally.
- Pricing
- Not published. Standard, Extended and Extensive pentest packages are quoted; attack surface tiers are sold by asset count.
Strengths
- Testing by in-house OSCP and CREST-certified testers, with no crowdsourced testers.
- Tests launch in 24 to 48 hours.
- Breach360, its AI pentester, asks for approval before any lateral movement or privilege escalation, and a certified pentester validates its findings.
- Reports mapped to SOC 2, PCI DSS, ISO 27001, HIPAA and HITRUST; Jira, ServiceNow, GitHub and Azure DevOps integrations.
Check before you buy
- Retest terms differ by page: the pricing page lists one or two free manual retests by package, the PTaaS page says unlimited. Ask which applies to your contract.
Astra Security
Scanner plus PTaaS- Best for
- Small teams that want a vulnerability scanner and a yearly pentest from one vendor at published prices.
- Pricing
- Published plan prices.
- Pentest Auto
- from $199/mo
- Pentest Expert (manual)
- $5,999/yr
- Pentest Enterprise
- from $9,999/yr
- DAST scanner
- from $69/mo
Strengths
- A full published price ladder across scanner, API, cloud and pentest plans.
- Manual pentests by certified testers, with CREST, PCI ASV and CERT-In compliant reports on Pentest Expert.
- AI agents with an independent AI validator on the automated tier, and a publicly verifiable pentest certificate.
Check before you buy
- The manual pentest takes 10 to 15 working days.
- Re-scans to verify fixes are capped: one on Pentest Auto, two on Expert, four on Enterprise.
- The pricing page shows two different prices for some plans; confirm at checkout.
AI-led testers
AI-led providers run the test with agents, and two of the three here publish a price per test. They differ on scope, and on who checks the agents' findings: a second agent, a person, or the same agent that found them. Helix Cyber is ours.
Helix CyberOurs
AI-led PTaaS- Best for
- SaaS teams that need an audit-ready test of a web app, API, network or cloud account within days, at a price known up front.
- Pricing
- Published fixed prices, confirmed before testing starts.
- Web app or API, black box
- $3,000/test
- Web app or API, gray box
- $4,000/test
- Web app or API, white box
- from $7,000/test
- External network
- from $3,000/test
- Internal network or cloud
- from $4,000/test
Strengths
- AI agents test in sandboxed desktops while people oversee the run. A second agent reproduces every finding, and anything it cannot reproduce never reaches your list.
- Report about 24 hours after scope, written authorization and test access are confirmed.
- Free, unlimited retests: mark a fix ready in the portal and it is attacked again.
- Open models on inference Helix controls. Your code, prompts and findings are not sent to OpenAI, Anthropic or another model provider.
- A full sample report is public: a gray-box test of OWASP Juice Shop with 67 reproduced findings.
Check before you buy
- No mobile app, social engineering, physical or red-team engagements.
- Agents do the testing, with people overseeing them. If a contract asks for named, individually certified testers, check with us before you buy.
- Larger or unusual scopes are priced after a scoping review, not from the list.
XBOW
Autonomous AI pentester- Best for
- Security teams that want autonomous, repeated testing of many web apps and APIs, with exploit proof for each finding.
- Pricing
- Not published. Usage-based, scoped to your environment; sold through cloud marketplaces too.
Strengths
- Every finding ships with the working exploit, request, response and reproduction steps.
- Free, unlimited re-tests after you deploy a fix.
- XBOW reports reaching #1 on HackerOne's leaderboard, with 1,500+ reproducible exploits on production targets.
- A separate safety model reviews each command before it runs; data residency in several regions.
Check before you buy
- Scope is web applications and APIs.
- Onboarding starts with a demo; XBOW's team helps set up credentials, network access and targets.
- “Findings are reviewed before you see them”; the page does not say whether a person or a system reviews them.
Intruder
Scanner and AI pentest- Best for
- Teams that want continuous external, cloud and web scanning, with an AI web app pentest on the same platform.
- Pricing
- Free plan and 14-day trial. Cloud and Pro plans are a base fee plus a per-target fee (amounts not verified: we could not read them from the pricing page). AI web app pentest $3,500 per test for platform customers, $4,000 standalone.
Strengths
- One platform for vulnerability scanning, attack surface monitoring, cloud checks and pentests, used by 3,000+ companies by Intruder's count.
- Emerging-threat scans run within hours of critical vulnerabilities going public.
- The AI pentest delivers a report within hours, includes unlimited retesting, and is refunded if your auditor rejects the report.
- Names its engines: Tenable, OpenVAS and Nuclei for infrastructure, a ZAP-based scanner for apps.
Check before you buy
- Intruder says “a human in the loop isn't necessary for web app pentesting anymore”; its agents validate findings, but no person does.
- The AI pentest covers web applications; internal target scanning needs the Pro or Enterprise plan.
Our guide to AI penetration testing covers what these tests can and cannot find yet.
Tools you run yourself
These are not companies that test for you. They are platforms your security team runs against its own networks as often as it likes. They complement an independent test; on their own they are a self-assessment.
Horizon3.ai NodeZero
Autonomous pentesting you run- Best for
- Security teams that want to run internal, external, cloud and Active Directory pentests themselves, as often as they like.
- Pricing
- Published list prices per asset. Core, Pro and Elite include unlimited tests for a year.
- Core
- $50/asset/yr
- Pro
- $65/asset/yr
- Elite
- $85/asset/yr
- Flex
- $15/asset/test
- WebApp Flex
- $10,000/asset/test
Strengths
- The clearest published pricing in this group, with a written definition of an asset.
- Internal tests run from a Docker host or OVA you set up in minutes; external tests run from Horizon3's cloud.
- Covers internal, external, AWS and Azure, Kubernetes, AD password audits and phishing-impact testing.
- NodeZero Federal is FedRAMP High Authorized, per Horizon3, and lists a FedRAMP package ID.
Check before you buy
- You operate it: someone on your team schedules runs and works the results.
- Support plans add 10% or 20% of the subscription, with minimums of $7,500 and $25,000 a year.
- The web app product is priced separately from the network subscriptions.
Pentera
Automated security validation- Best for
- Large security teams that want to validate internal networks, Active Directory, external assets and cloud continuously, on their own schedule.
- Pricing
- Not published.
Strengths
- One product family for internal networks (Core), external attack surface (Surface), cloud and web apps and APIs.
- Safe, assumed-breach tests without prior credentials, including offline cracking of AD password hashes.
- Pentera says it has been proven across 1,300+ customer environments over 10 years.
- A managed option (SECTOR11) for teams that want Pentera's experts to run the tests.
Check before you buy
- The deployment model is not described on the pages we read.
- Pentera states it does not certify compliance or claim FedRAMP authorization.
Comparing these two? See Pentera alternatives.
How to choose a penetration testing company
Work through these in order. The first two usually remove most of the list.
- Why you need the test. An auditor, a customer questionnaire, a release, or a board question. Compliance-driven tests need a report the auditor accepts; see our guides to SOC 2, PCI DSS and ISO 27001 pentest requirements.
- What is in scope. Web apps and APIs are covered by almost everyone. Mobile, hardware, OT, physical and social engineering narrow the list to consultancies and a few platforms.
- Who tests, and who checks. Employed consultant, contract researcher, or AI agent; and whether a person or a second check confirms each finding before you see it.
- Access level. Gray-box tests with real accounts find the access-control and tenant-isolation bugs auditors and enterprise customers care about. Our black, gray and white box comparison explains the trade-off.
- Retest terms. On this page they range from one re-scan within 30 days to unlimited retests for 12 months. Fixes without a retest leave findings open in the report.
- A sample report. Read one before you buy. Ours is public: a gray-box test of OWASP Juice Shop.
- Where your data goes. If an AI provider tests your app, ask which model provider processes your code, traffic and findings.
Frequently asked questions
What does a penetration testing company do?
It attacks your systems with your written permission to find weaknesses a real attacker could use, proves which ones are exploitable, and reports each finding with evidence, a severity rating and a fix. Most also retest after you fix the issues, so the report shows what was resolved.
How much does a penetration test cost in 2026?
Among providers that publish prices, AI-led tests of one web app or API start around $3,000 to $4,000 (Helix Cyber, Intruder, Cobalt's Autonomous Pentest offer). Human-led tests start around $6,000 to $11,000 (Astra's Pentest Expert, Synack's Standard pentest, Software Secured). Consultancies such as Bishop Fox, NCC Group and Rapid7 quote per engagement. Scope, access level and retest terms move the price more than the provider does.
What is the difference between PTaaS and a traditional pentest company?
A traditional consultancy runs a project: scoping calls, a test window, then a PDF. Penetration testing as a service (PTaaS) puts ordering, findings, retests and reports in an online platform, often with testers drawn from a vetted community. The testing itself can be just as deep; the difference is how you buy and follow it.
Are AI penetration tests real pentests?
The good ones exploit and prove each finding. That proof is what separates a pentest from a scan. Ask how a finding is confirmed before it is reported: by a person, by a second independent check, or not at all. Also ask what the AI cannot test yet. Synack, for example, states that its AI pentester does not yet support MFA logins.
How long does a penetration test take?
AI-led tests report within hours to about a day once scope and access are ready. Human-led PTaaS tests usually start within one to three business days and run for days to two weeks; Astra, for example, quotes 10 to 15 working days for its manual pentest. Consultancy engagements are often booked weeks ahead.
How do I check that a penetration testing company is legitimate?
Ask for a sample report, the names of the certifications held by the people or company doing the work, the rules of engagement you will sign, and proof of insurance. A legitimate provider will also verify that you own or are authorized to test every asset in scope before it starts.
Sources, checked November 5, 2026
Each vendor's claims are quoted or summarised from the pages below. Customer counts, ratings and percentages are the vendor's own figures; we have not audited them. “Not published” means we found no price on the vendor's site. “Not verified” means the page draws its prices with a script we could not read, so a price may exist that we could not confirm. If something here is wrong or out of date, tell us and we will correct it.