XBOW
XBOW · XBOW: Autonomous AI Pentesting for Web Apps and APIs, With a Working Exploit Behind Every Finding
XBOW is an autonomous offensive security platform that pentests web applications and their APIs with coordinated AI agents, reports only vulnerabilities it can reproduce with a working exploit, and retests each fix after you ship it. It suits AppSec and enterprise security teams that want to test far more applications than manual pentests reach, with auditor-ready reports, sales-led usage-based pricing and purchase through the AWS, Google Cloud, Oracle and Microsoft marketplaces.
Best for
Enterprise and mid-market security teams that own more web applications than they can pentest by hand and want exploit-validated findings, fix retesting and audit evidence across the whole portfolio, bought through sales or a cloud marketplace
Not ideal for
Teams that need to test mobile apps, desktop or compiled software, internal apps that cannot be exposed to XBOW's IP addresses, or standalone APIs without a web front end; buyers who want a published price, a free trial or an open-source tool they can run themselves
Who it's for
Application security and offensive security teams, CISOs at mid-market and enterprise companies, and security engineering teams that need continuous or on-demand pentests across many web applications
Capabilities
- Autonomous pentests of interactive web applications and their APIs, in production or staging, unauthenticated or signed in as a test account
- Coordinator-plus-agents architecture: a persistent coordinator maps the target and directs many short-lived attack agents in parallel, each retired after its mission
- Validators confirm each exploit is reproducible before a finding is shown, with deterministic, non-AI checks for generally available vulnerability classes; preview attack methods are opt-in and labeled, and private-preview findings are triaged by XBOW's security team
- Authentication by username and password, TOTP or email one-time codes, SSO through an identity provider such as Okta, magic links, GitHub or Microsoft social login, bearer tokens and HTTP basic auth, with a sequential mode for apps that cannot handle concurrent sessions
- Gray-box context: upload a source-code and documentation archive (up to 5 GB), OpenAPI specs and plain-language assessment guidance to focus testing
- Three impact-demonstration levels (Off, Moderate, Deep) that control how far XBOW escalates a confirmed finding, plus canary tokens you plant to prove business-logic, SQL injection, file-read and RCE findings
- Safety layers: a separate AI safety judge and deterministic rules check every command, first-party and third-party domains are scoped automatically, sensitive URLs can be blocked or auth-only, and an assessment pauses if the target becomes unhealthy
- PDF pentest report, executive summary and signed letter of attestation generated automatically after each successful assessment, with a new report version each time a fix is verified
- Retest assessments that replay the original exploit and try bypasses to confirm a fix holds
- Autonomous Exposure Management: managed discovery of your external application estate and parallel black-box testing across the approved list
- Versioned REST API for assets, assessments, findings and reports, with signed webhooks for finding, assessment and asset changes
- Jira app that turns validated findings into work items with optional status writeback, and Microsoft Sentinel connector plus two Security Copilot agents to query findings and launch pentests in natural language
- Role-based access, an audit log, and SSO via SAML 2.0 or OIDC with SCIM provisioning (SSO and SCIM on Enterprise)
Limitations
- Only interactive web applications and their APIs are supported. XBOW's docs say it does not test APIs that have no web application front end, and support for other asset types, such as mobile apps or compiled software, is described only as on the roadmap
- Targets must be reachable by XBOW from the internet, with XBOW's IP addresses allowlisted in any firewall or WAF; apps behind a VPN, unstable test environments and end-of-life infrastructure give limited or unreliable results
- Some authentication flows are unsupported: SMS-based MFA, Google social login, CAPTCHA challenges, account registration and partially onboarded accounts
- No public price list: the pricing page asks for a quote scoped to your environment, and pricing is usage-based
- Users join through an organization rather than signing up individually, so there is no documented free trial or self-serve sign-up
- The standard platform is hosted in the United States. Data residency in Australia, Canada, Europe, Japan and Singapore is in private preview, the region choice is permanent, and outside Europe, AI model prompts may still be processed outside the chosen region
- The documented product is a multi-tenant SaaS; XBOW's docs only point customers who need to run it in their own environment to sales for self-hosted deployment, and the platform is closed source
- The Jira, Microsoft Sentinel and Security Copilot integrations and SSO are labeled public preview in the docs, and the Security Copilot agents consume Microsoft Security Compute Units
Use cases
- Running an authenticated pentest of a customer-facing web app before a release and handing auditors the report and letter of attestation
- Testing a large external application estate, including inherited or forgotten apps, from an attacker's starting point with no credentials
- Triggering a pentest from CI or a release scheduler through the API and gating the release on validated findings
- Verifying that a fix actually closed a vulnerability by replaying the original exploit and trying bypasses
- Sending only exploit-validated findings into Jira or Microsoft Sentinel so remediation starts from proof instead of scanner output
Our take
XBOW's pitch is proof over volume: a finding only reaches you after a non-AI validator reproduces the exploit, and the trace shows exactly how it got there. That is the right answer to scanner fatigue, and the docs back it with unusually concrete safety controls and a clear split between generally available and preview attack methods. The fit is narrower than the marketing suggests. XBOW tests internet-reachable web apps with a browser front end, so mobile, compiled software, internal-only apps and standalone APIs fall outside it, and buying is a sales conversation with no public price. For an enterprise with hundreds of under-tested apps, Exposure Management and the API are the main draw. For a startup that needs one annual pentest, a fixed-price service such as Fleuret or Hadrian's Nova, or an open-source agent such as Strix, may be simpler to buy.
Who should use it
Security teams responsible for many internet-facing web applications that want continuous, exploit-validated testing with retests and audit evidence, especially organizations that can buy through a cloud marketplace and want findings flowing into Jira, Microsoft Sentinel or their own tooling through the API.
Who should skip it
Teams that need mobile, desktop, internal-network or standalone API testing, organizations that cannot use a US-hosted SaaS while other residency regions are in private preview, buyers who want transparent self-serve pricing or a free trial, and anyone without authorization to test the target.
Strengths
- Findings come with the request, response and full attack trace, and generally available vulnerability classes are confirmed by deterministic, non-AI validators
- Free, unlimited retests replay the original exploit and try bypasses after a fix
- Scales from one app to portfolio-wide testing, with a managed Exposure Management offering for large estates
- Detailed public documentation of scope rules, safety layers, impact levels and supported authentication
- Pentest report, executive summary and letter of attestation generated automatically for auditors
- Versioned REST API, webhooks, Jira and Microsoft Sentinel and Security Copilot integrations, and purchase through four cloud marketplaces
Weaknesses
- No published pricing, free trial or individual sign-up
- Limited to internet-reachable web apps with a browser front end; standalone APIs, mobile and compiled software are out of scope for now
- Some common login setups (SMS MFA, Google sign-in, CAPTCHA) are not supported
- US-hosted by default, with other data residency regions still in private preview and no documented self-hosted option
- Several integrations and SSO are still labeled public preview
XBOW pricing
XBOW platform
Custom
- Usage-based pricing scoped to your environment
- Quote on request
- Available through AWS, Google Cloud, Oracle and Microsoft marketplaces
- Standard support included with every subscription
Autonomous Exposure Management
Custom
- Managed discovery of your external application estate
- Parallel black-box testing across the approved list
- Run by XBOW field engineers and specialists
Note: XBOW's pricing page lists no prices and describes usage-based pricing that scales with coverage rather than a fixed annual engagement; homepage copy says retests are free and unlimited. Pentest On-Demand was announced in November 2025 as self-service from $4,000, but the current site routes buyers to a quote or demo. Advanced and Strategic support tiers are optional upgrades.
Where XBOW excels
Covering the applications manual pentests never reach
Exposure Management discovers the external app estate, you approve the target list, and agents test every app in parallel from an unauthenticated starting point, so lower-priority apps get tested alongside the crown jewels.
Compliance pentest evidence for SOC 2 or ISO 27001
Each successful assessment produces a full pentest report, a leadership summary and a signed letter of attestation that can go to auditors or customers without exposing exploit details.
Pre-release security gate
The API can trigger an assessment on merge or before deploy, and webhooks report validated findings, so a release can wait on proven issues rather than unverified scanner output.
XBOW vs. competitors
XBOW vs. Hadrian
Hadrian combines agentic attack-surface management (Atlas) with on-demand AI pentests of a single web app (Nova) at a published per-test price, with human-reviewed, compliance-mapped reports. XBOW tests web apps and APIs with exploit validators and free retests, scales to portfolio-wide testing through its API and Exposure Management, and quotes prices on request.
XBOW vs. Neo
Neo is ProjectDiscovery's self-serve AI security platform with per-seat pricing, covering pentesting alongside PR security review, vulnerability report triage and internal red teaming. XBOW concentrates on exploit-validated web app and API pentesting at enterprise scale, sold through sales and cloud marketplaces.
XBOW vs. Fleuret
Fleuret is a Paris-based AI pentesting service hosted in France that combines agents with human offensive expertise, with a one-off pentest at €4,000 per web app and published yearly credit plans. XBOW is a US-hosted autonomous platform aimed at continuous and portfolio-wide testing, with custom usage-based pricing and data residency regions in private preview.
XBOW vs. Strix
Strix is an Apache-2.0 open-source AI pentesting tool you run in your CLI or CI with your own model key, plus a paid cloud. XBOW is a closed, managed platform with validators, retests, compliance reports and enterprise controls such as SSO and SCIM.
XBOW vs. Shannon
Shannon is Keygraph's AGPL-3.0 white-box pentester that reads your source code and tests the running app, run locally or in CI. XBOW is a hosted service that can test with or without source code and adds portfolio-scale orchestration, reporting and integrations, sold as a managed service rather than a tool you run yourself.
Frequently asked questions
What is XBOW?
XBOW is an autonomous offensive security platform. A coordinator directs many short-lived AI agents to map and attack a web application and its API, independent validators confirm each exploit, and the platform reports only reproducible findings with evidence and remediation guidance, then retests fixes.
How much does XBOW cost?
XBOW does not publish prices. Its pricing page describes usage-based pricing scoped to your environment and asks you to request a quote, and it can be bought through the AWS, Google Cloud, Oracle and Microsoft marketplaces. When Pentest On-Demand was announced in November 2025, it was priced from $4,000; check current terms with XBOW.
What can XBOW test?
Interactive web applications and their APIs, authenticated or unauthenticated, in production or staging. The target must be reachable by XBOW from the internet, allowlist XBOW's IP addresses in any firewall or WAF, and work in Chrome. XBOW's docs say it does not test APIs without a web application front end, and other asset types are on its roadmap.
Does XBOW do black-box or gray-box testing?
Both. It can test from the outside with no credentials, which is how Autonomous Exposure Management starts, or sign in with test accounts and use uploaded source code, documentation and API specs for gray-box testing.
Is XBOW safe to run against production?
XBOW is designed for it: every command is checked by a separate AI safety judge and fixed rules, exploits are demonstrated with benign techniques such as a one-time sleep for SQL injection, you control scope, blocked URLs and impact level, and the assessment pauses if the target becomes unhealthy. Every action is logged.
Can XBOW be self-hosted?
Not as a self-serve option. The documented platform is SaaS hosted in the United States, with data residency regions in Australia, Canada, Europe, Japan and Singapore in private preview. XBOW's docs tell organizations that need to run it entirely in their own environment to contact sales about self-hosted deployment, but publish no details.
XBOW vs Hadrian: what is the difference?
Both sell agentic pentesting with compliance-ready reports. Hadrian pairs per-test pentests of a single app (Nova, from a published per-test price) with continuous attack-surface management (Atlas). XBOW focuses on exploit-validated testing of web apps and APIs at portfolio scale, with an API, retests and a managed Exposure Management engagement, and quotes prices on request.
Integrations & fit
Alternatives to consider
Hadrian Security B.V.
CISO-led security teams at mid-market and enterprise organizations (Hadrian names 300 to 10,000 employees) with a large, changing internet-facing estate across cloud and on-premises, who want validated external exposures instead of scanner noise and pentest evidence for NIS2, DORA, ISO 27001 or SOC 2
Paid
ProjectDiscovery
AppSec and product security teams, MSSPs and bug bounty hunters who want an agent to pentest running apps, review pull requests and reproduce incoming vulnerability reports with proof, on a platform that keeps context between runs
Freemium
Fleuret AI (FLEURET AI SAS)
European companies of roughly 50 to 1,000 people with a real web and API attack surface (SaaS, fintech, healthtech) and a NIS2, DORA, SOC 2 or ISO 27001 deadline, especially those that want pentest data and model inference kept on EU infrastructure
Freemium
Strix (OmniSecure, Inc.)
Engineering and AppSec teams that want open-source, AI-driven pentesting of their own apps and APIs in the CLI or CI, with an option to move to a managed platform
FreemiumAbout XBOW
XBOW runs a pentest without a human driving the attack. You add an asset (an interactive web application and its API), optionally supply test-account credentials, API specs, documentation or a source-code archive for gray-box testing, and set the scope, protected URLs, impact level and testing window. A preflight check validates the target, credentials and domains, then a persistent coordinator maps applications, endpoints and auth flows and dispatches short-lived attack agents, each with one objective and fresh context, that work in parallel with an offensive toolkit and a headless browser. Validators confirm an exploit is reproducible before a finding surfaces: findings in generally available vulnerability classes are checked by deterministic, non-AI validators, attack methods in public or private preview are opt-in and labeled, and private-preview findings are triaged by XBOW's security team. Each finding comes with the request, response, full attack trace and remediation guidance, and every successful assessment produces a PDF pentest report, an executive summary and a signed letter of attestation that XBOW says meet SOC 2 and ISO 27001 pentest requirements. After a fix, a retest replays the original exploit and tries bypasses. Since September 2026 XBOW also sells Autonomous Exposure Management, a managed engagement in which its field engineers discover your external application estate and XBOW tests the approved list from an attacker's starting point, with no source code or credentials, across dozens to thousands of apps. A versioned REST API, webhooks, a Jira app and Microsoft Sentinel and Security Copilot integrations push this into existing workflows. XBOW's public reputation rests on results it publishes itself, including reaching #1 on the HackerOne leaderboard in June 2025. The tradeoffs: scope is limited to internet-reachable web apps with a browser front end (standalone APIs, mobile apps and compiled software are not supported yet), pricing is a custom quote, and it is a closed, US-hosted SaaS by default, with other regions in private preview and self-hosting mentioned only as a sales conversation.
Updates from XBOW
On October 13, 2026, XBOW published how its researchers, working with AI agents on a publicly available OpenAI model, built a Chrome full-chain exploit: the awarded chain of three Chrome bugs, from a web page to full code execution on Windows, was built by one researcher and the agents in six days and earned Google's first $250,000 Chrome full-chain reward of 2026. This was a research project rather than a feature of the web pentesting product; XBOW says it is now developing an autonomous product for compiled software. A follow-up webinar ran on October 28, 2026.
XBOW added Australia, Canada and Japan data residency regions on October 14, 2026, alongside Europe and Singapore. Data residency is in private preview on multi-tenant SaaS and is provisioned through sales.
On September 24, 2026, XBOW introduced Autonomous Exposure Management, a managed engagement that discovers an organization's external application estate and tests every approved app in parallel from an attacker's starting point, with no source code or credentials, reporting only findings it can reproduce.
On March 23, 2026, XBOW announced an integration with Microsoft Sentinel and Microsoft Security Copilot: a Sentinel connector imports findings for alerting and querying, and two Security Copilot agents analyze findings and launch pentests from natural-language prompts.
XBOW's Public API launched in public preview on February 1, 2026, with programmatic control of assets, assessments, findings and reports and webhooks for change events. The docs now list dated stable API versions.
On November 13, 2025, XBOW announced Pentest On-Demand, a self-service automated pentest with results within five business days, retesting and compliance-ready documentation, priced from $4,000 at launch. XBOW's current pricing page asks for a quote instead of listing this price.
Are you the founder? Claim this listing →