# Xalgorix > Hosted autonomous AI pentesting. Runs a 22-phase offensive methodology against web apps and APIs and ships verified, evidence-backed findings on demand. Xalgorix is a SaaS that turns the `xalgo` offensive engine into a click-to-scan service. Users pick a target, choose a scan mode (single / wildcard), and receive live phase progress, severity-rated findings with evidence, and a branded PDF report. Pricing is credit-based with monthly subscriptions and one-time top-up packs. The engine is open source (Apache-2.0) at https://github.com/xalgorix/xalgorix and can be self-hosted; [Cloud vs self-hosted](https://www.xalgorix.com/hosted-vs-self-hosted) compares the two. ## Pages - [Home](https://www.xalgorix.com/): Product overview, 22-phase methodology, scan modes, and a sample finding. - [Methodology](https://www.xalgorix.com/methodology): The 22 phases Xalgorix runs on every engagement. - [Sample reports](https://www.xalgorix.com/sample): Two real Xalgorix reports from live targets, with severity, CVSS and downloadable PDFs. No login. - [Pricing](https://www.xalgorix.com/pricing): Subscription plans (Hacker $49/mo, Pro $119/mo, Max $279/mo) and one-time credit packs from $10. - [Use cases](https://www.xalgorix.com/use-cases): How developers, AppSec teams, founders, MSSPs, compliance teams and bug bounty hunters use Xalgorix. - [Bug bounty use case](https://www.xalgorix.com/use-cases/bug-bounty): Running Xalgorix within a bug bounty program's rules: identifying headers, rate limits, scope and exclusions, and what it can't enforce. - [Pentesting for startups](https://www.xalgorix.com/use-cases/startups): Testing before launch and after each release without a security team, sharing evidence with customers, and what SOC 2 and PCI DSS ask for. - [Cloud vs self-hosted](https://www.xalgorix.com/hosted-vs-self-hosted): Run the open-source engine on your own infrastructure, or let Xalgorix Cloud run models, compute and out-of-band listeners. - [GitHub App](https://www.xalgorix.com/github-app): Free security review comment on every pull request. No workflow file, API key or account. - [Integrations](https://www.xalgorix.com/integrations): Cloud: free GitHub App, scan Action, REST API, signed webhooks, Slack and Discord alerts, scheduled scans. Self-hosted engine: LLM providers, proxy, Caido, AgentMail, Discord and Telegram settings. - [Blog](https://www.xalgorix.com/blog): Exploit-verification guides, AppSec field notes, and weekly threat digests. RSS feed: [/rss.xml](https://www.xalgorix.com/rss.xml). - [Docs](https://www.xalgorix.com/docs): Install and run the self-hosted CLI: configuration, flags, environment variables and local API. - [API docs](https://www.xalgorix.com/docs/api): Public REST API: start scans, poll status, pull findings, plus webhooks, status badges, and the GitHub Action. - [Sign up](https://www.xalgorix.com/signup): Create an account (credit packs from $10). ## Comparisons - [Open-source AI pentesting tools compared](https://www.xalgorix.com/compare): Shannon vs Strix head to head, plus license, approach, hosted pricing and model requirements for all four tools, and FAQs on Xalgorix alternatives and which AI pentesting tools you can self-host. - [Xalgorix vs Strix](https://www.xalgorix.com/compare/strix): License, hosted pricing, setup, PR reviews and when to choose each. - [Xalgorix vs Shannon](https://www.xalgorix.com/compare/shannon): What Keygraph is; Shannon open source is white-box and Keygraph Pro adds black-box; Xalgorix tests the running app. - [XBOW alternatives](https://www.xalgorix.com/compare/xbow-alternatives): Eight options: open-source agents you can self-host (Xalgorix, Strix, Shannon, PentestGPT) and commercial platforms (RunSybil, Escape, NodeZero, Pentera). - [Xalgorix vs XBOW](https://www.xalgorix.com/compare/xbow): Enterprise, quote-based platform vs open source with self-serve Cloud. - [Xalgorix vs PentestGPT](https://www.xalgorix.com/compare/pentestgpt): A research-born, open-source-only agent strong on CTFs vs exploit-verified web and API pentests, self-hosted or hosted. ## Guides - [Writing bug bounty reports triagers can act on](https://www.xalgorix.com/blog/bug-bounty-report-writing-guide): How to write a bug bounty report that gets triaged: titles, attack scenario, reproducible steps, defensible impact, CVSS reasoning and AI-assisted checks. - [Security regression tests that catch incomplete fixes](https://www.xalgorix.com/blog/security-regression-tests-vulnerability-fixes): Turn each fixed vulnerability into tests that fail CI: invariant-based pytest checks, a route-coverage guard and a Nuclei job with a control run. - [SSRF filter bypass through URL parser confusion](https://www.xalgorix.com/blog/ssrf-filter-bypass-url-parser-confusion): How backslashes, userinfo, hex and octal IPs and IPv4-mapped IPv6 slip past SSRF filters, with reproducible tests, real CVEs and a validator that holds. - [npm and PyPI trusted publishing: hardening the release job](https://www.xalgorix.com/blog/trusted-publishing-npm-pypi-release-hardening): Trusted publishing removes npm and PyPI tokens but leaves workflow compromise open. Environments, tag rules, staged publishing and provenance close the gap. - [ruby-saml SAML signature bypasses and parser differentials](https://www.xalgorix.com/blog/ruby-saml-signature-bypass-parser-differential): How five ruby-saml CVEs let one IdP-signed document forge any SAML login, from REXML vs Nokogiri to void canonicalization, and the SP fixes that close them. - [Scoping AI agent and MCP tool permissions](https://www.xalgorix.com/blog/ai-agent-tool-permissions-least-privilege): Why the GitHub and Supabase MCP leaks were permission failures, and five controls that limit an injected AI agent, from scoped tokens to database roles. - [Stop NoSQL operator injection in Node.js and Mongoose](https://www.xalgorix.com/blog/nosql-operator-injection-mongodb-mongoose): How NoSQL operator injection works in Express and Mongoose, what sanitizeFilter covers, its two 2026 bypasses, and the input validation that stops it. - [Serving user uploads safely: domains, headers, sandboxing](https://www.xalgorix.com/blog/serve-user-uploads-safely-sandbox-domain): Sandbox domains, __Host- cookies, nosniff, Content-Disposition, CSP sandbox and signed URLs: how to serve user-uploaded files so a hostile one stays harmless. - [Automated penetration testing: types, limits, how to choose](https://www.xalgorix.com/blog/automated-penetration-testing): What automated penetration testing is, the four kinds of tool sold under the name, what they miss, what PCI DSS and SOC 2 expect, and how to evaluate one. - [DAST vs AI Pen Testing vs Vulnerability Scanners Compared](https://www.xalgorix.com/blog/ai-pentesting-vs-dast-vs-vulnerability-scanners): How AI pentesting agents, DAST and vulnerability scanners differ in coverage, false positives, cost, CI/CD fit and production safety, and when to use each. - [OAST explained: reading DNS and HTTP callbacks](https://www.xalgorix.com/blog/out-of-band-testing-oast-callbacks): How out-of-band testing works, what a DNS-only vs HTTP callback means, labelling payloads, delayed and third-party pingbacks, and self-hosting interactsh. - [JWT jwk, jku and kid header injection attacks](https://www.xalgorix.com/blog/jwt-header-injection-jwk-jku-kid): How jwk, jku, x5u, x5c and kid header injection let attackers pick the key that verifies a JWT, from CVE-2018-0114 to Authlib’s 2026 bug, and how to stop it. - [How Copilot and Cursor agents got RCE via config files](https://www.xalgorix.com/blog/ai-coding-agent-config-file-rce-copilot-cursor): CVE-2025-53773, -54135, -54136 and -59944: AI coding agents that edited their own config to run code. What went wrong in each, and the controls that hold. - [Strict CSP rollout with nonces and Trusted Types](https://www.xalgorix.com/blog/strict-content-security-policy-nonces-trusted-types): Deploy a strict Content Security Policy: nonce vs hash, strict-dynamic, a report-only rollout, Trusted Types for DOM XSS, and what it still misses. - [pull_request_target pwn requests: what GitHub’s fixes miss](https://www.xalgorix.com/blog/github-actions-pull-request-target-pwn-requests): GitHub’s checkout v7, read-only cache and default-branch changes block some pwn requests. Four GitHub Actions routes they leave open, and how to audit. - [A signed token is not proof: verifying OIDC flaws](https://www.xalgorix.com/blog/verifying-openid-connect-relying-party-flaws): A validly signed ID token can still be forged, replayed, minted for another app, or from the wrong provider. How to verify OpenID Connect sign-in flaws. - [Verifying SQL injection: proof, not a DB error](https://www.xalgorix.com/blog/verifying-exploitable-sql-injection): A DB error or a slow response is a lead, not proof of SQL injection. How to tell a real, exploitable SQLi from an innocent blip, confirm it safely, and fix it. - [Verifying exploitable XSS: proof, not a popup](https://www.xalgorix.com/blog/verifying-exploitable-xss): An alert(1) proves a sink ran, not that the XSS is exploitable. How to tell self-XSS and neutralized payloads from impact in a victim's session, and fix it. - [SSRF DNS rebinding: why filters fail, how to pin](https://www.xalgorix.com/blog/ssrf-filter-dns-rebinding-resolve-and-pin): An SSRF filter that validates one DNS lookup, then lets the client connect on another, falls to DNS rebinding. Why it happens and the resolve-and-pin fix. - [Verifying web race conditions: proof, not luck](https://www.xalgorix.com/blog/verifying-web-race-conditions): A limit that breaks once could be luck, not a race condition. How to tell a timing fluke from a real concurrency bug, confirm it safely, and fix the window. - [An accepted upload is not proof: verifying file upload](https://www.xalgorix.com/blog/verifying-exploitable-file-upload): A server accepting or storing your file is not proof of a bug. How to verify an exploitable file upload — RCE or stored XSS — and the fixes that hold. - [A 9.8 is not a priority: CVSS vs exploitability](https://www.xalgorix.com/blog/cvss-score-is-not-a-priority): A CVSS 9.8 measures worst-case severity, not your risk. How to rank vulnerabilities by exploitability and reachability using EPSS, KEV and SSVC. - [A missing token is not proof: verifying CSRF](https://www.xalgorix.com/blog/verifying-csrf-vulnerability): A missing or unvalidated CSRF token is not proof of a vulnerability. How to verify a real cross-site request forgery in the browser, and the fixes that hold. - [Verifying XXE: proving external entity injection](https://www.xalgorix.com/blog/verifying-xxe-injection): A parsed entity or an XML error is not proof of XXE. How to verify a real external entity injection — file read or SSRF — safely, plus the fix that holds. - [Verifying prompt injection in LLM applications](https://www.xalgorix.com/blog/verifying-prompt-injection-llm-applications): A model echoing an injected instruction is not proof of a vulnerability. How to verify a real prompt injection — a trust boundary actually crossed — safely. - [Verifying broken function level authorization (BFLA)](https://www.xalgorix.com/blog/verifying-broken-function-level-authorization): A reachable admin function and a 200 are not proof of BFLA. How to verify a real broken-function-level-authorization escalation safely, and the fixes that hold. - [A reflected Origin is not proof: verifying CORS](https://www.xalgorix.com/blog/verifying-exploitable-cors-misconfiguration): A site echoing your Origin in Access-Control-Allow-Origin looks like a bug but often proves nothing. How to verify an exploitable CORS misconfiguration safely. - [A 302 is not proof: verifying OAuth account takeover](https://www.xalgorix.com/blog/verifying-oauth-redirect-uri-account-takeover): A 302 bouncing an OAuth login to your server looks dramatic but proves little. How to verify a redirect_uri account takeover safely, and the fixes that hold. - [How to confirm HTTP request smuggling safely](https://www.xalgorix.com/blog/confirming-http-request-smuggling-safely): A timing blip is not proof of HTTP request smuggling. How to confirm a desync safely, without poisoning other users, and the fixes that hold. - [{{7*7}} is not RCE: proving template injection](https://www.xalgorix.com/blog/proving-server-side-template-injection): A template that evaluates {{7*7}} to 49 proves evaluation, not RCE. How to tell detection from an exploitable sandbox escape and actually verify SSTI. - [Verifying IDOR and BOLA: a 200 is not proof](https://www.xalgorix.com/blog/verifying-idor-bola-access-control): Changing an ID and getting a 200 isn't proof of IDOR. Confirm broken object level authorization with a two-account test and scope it without touching real data. - [GraphQL security beyond introspection: what holds](https://www.xalgorix.com/blog/graphql-security-introspection-batching-dos): Auth on, introspection off, a rate limit set — yet a GraphQL login still brute-forces. How aliasing and batching beat throttles, and the limits that hold. - [How to Verify SSRF Is Actually Exploitable](https://www.xalgorix.com/blog/verifying-ssrf-exploitability): Finding SSRF is easy; proving it is hard. Confirm it with out-of-band detection, see why blocklists and DNS rebinding fail, and the controls that stop it. - [When an AI Pentester Escapes the Lab: Agent Safety Lessons](https://www.xalgorix.com/blog/when-an-ai-pentester-escapes-the-lab): Anthropic disclosed AI agents reaching real systems during CTF evals. Lessons on scope, egress and credentials for safe autonomous pentesting. - [What It Costs to Self-Host an AI Pentester](https://www.xalgorix.com/blog/what-self-hosting-an-agentic-pentester-costs): Self-hosting Xalgorix is free, but LLM tokens, infra and upkeep are not. A cost model for agentic scans, and how Cloud credits compare. - [Xalgorix Hands-On Guide: Setup, Scans & Trade-offs](https://www.xalgorix.com/blog/xalgorix-review): A hands-on walkthrough of Xalgorix by its founder: install paths, the 22-phase methodology, verification, strengths and limits. - [Xalgorix Now Reviews Pull Requests: GitHub App Launch](https://www.xalgorix.com/blog/xalgorix-reviews-your-pull-requests): Install the free Xalgorix GitHub App for automatic security reviews on every pull request. No workflow file or API key. Comment to re-run. - [Exploit-Verified Pentesting: Why Detection Isn't Proof](https://www.xalgorix.com/blog/detection-is-not-proof): Most scanners flood you with maybes. See how Xalgorix exploits each finding and re-verifies it independently, so every result is evidence. ## API & integrations - REST API (Bearer `xlg_live_…` keys): `POST /api/public/v1/scans` to start a scan, `GET /api/public/v1/scans?id=` for status, `GET /api/public/v1/findings?scan_id=` for findings. 30 scans/min per key. - Webhooks: register HTTPS endpoints to receive a signed `scan.completed` event (HMAC-SHA256 over `${timestamp}.${body}`, `X-Xalgorix-Signature` header). - Status badge: `GET /api/public/v1/badge?slug=` returns an SVG for embedding in a README (params: `metric`, `label`). No auth. - GitHub Action: `xalgorix/scan-action@v1` (https://github.com/xalgorix/scan-action) runs a scan in CI and can fail the build on findings above a severity threshold. ## Optional - [Changelog](https://www.xalgorix.com/changelog): Open-source engine releases and Xalgorix Cloud updates. - [Security](https://www.xalgorix.com/security): How customer data and scan evidence are handled. - [Vulnerability disclosure](https://www.xalgorix.com/vdp): Scope, rules and safe harbor for reporting Xalgorix vulnerabilities. - [Status](https://www.xalgorix.com/status): Live service checks. - [About](https://www.xalgorix.com/about): Who builds Xalgorix and why. - [Optimizely Hijack, Axis RCE, nginx RCE: AppSec Digest](https://www.xalgorix.com/blog/weekly-appsec-threat-digest-nov-13-20-2026): Nov 13–20, 2026: Optimizely scripts hijacked on 178 orgs, Apache Axis RCE exploited via Confluence, a vCenter zero-day, nginx HTTP/2 RCE and Java4Shell.