RX VeX 2.1R is a large, low-latency cyber intelligence model. It fuses a live fetch of the target with a seven-lane audit graph, then writes a /100 report grounded in evidence — not guesses.
Current reasoning core — large context, tight decoding, cyber-tuned.
Parallel audit lanes scored on every authorized target.
Residual risk score — every report is graded /100.
Existence and fingerprint only. No exploit payloads. Ever.
Not a generic chatbot pasted onto headers. A constrained reasoning stack built for authorized website intelligence.
2.1R never scores a page it did not see. The first pass is a real HTTP fetch — status, headers, body, cookies, redirects — then the model reasons over that packet, not a training memory of the domain.
Seven specialist heads run in one pass: security, recon, SEO, performance, privacy, accessibility, technology. Findings are fused into a single residual score so a missing CSP cannot hide behind a perfect title tag.
Low-temperature, evidence-first generation. 2.1R is steered to cite what was fetched, rank what to fix, and refuse exploit recipes. Tight sampling keeps reports stable across rescans.
The core is fed a Hunt graph, not a homepage. Certificate Transparency names, Wayback CDX history, in-scope crawl, live host probes, and signature fan-out all land in the same context window.
Paste a URL. 2.1R scores each aspect of the target, with evidence you can verify. Examples below are the kind of finding a real scan returns.
CSP, HSTS, X-Frame-Options, Referrer-Policy, cookie flags (HttpOnly, Secure, SameSite), certificate and redirect hygiene. Missing controls are treated as ground truth from the live response.
MEDIUM Missing Content-Security-Policy target https://northstar.app evidence GET / → 200 · no CSP header impact injected scripts run in the victim origin fix add_header Content-Security-Policy "default-src 'self'" always;
Certificate Transparency subdomains, Wayback historical URLs, in-scope crawl, live host probes, JS param harvest, exposed paths, dangling CNAMEs. Fan-out signatures hit more than the URL you typed.
target northstar.app Subdomains 14 found api.northstar.app staging.northstar.app cdn.northstar.app Exposed paths /.git/HEAD [403] /admin [302] Wayback 214 historical URL(s) Live hosts 9 probed · Fastly
Title, description, canonical, robots.txt, sitemap.xml, Open Graph. 2.1R flags pages that search engines cannot index — or that leak staging into public metadata.
LOW robots.txt missing or empty target https://northstar.app evidence GET /robots.txt → 404 impact crawlers guess paths; staging may get indexed fix publish Disallow rules for /admin /api/internal
Compression, caching headers, content-type, page weight signals. Not a lab Lighthouse run — a first-packet read of how the origin actually serves the document.
LOW HTML served without compression target https://northstar.app · 1.8 MB document evidence Content-Encoding absent on text/html impact slower TTFB on mobile / weak links fix gzip / brotli on text, json, svg, js
Third-party scripts, Permissions-Policy, Referrer-Policy, cookie surface. The report covers more than firewalls — what the page phones home to, and what the browser is allowed to do.
MEDIUM Permissions-Policy missing target https://northstar.app evidence 3 third-party scripts · no Permissions-Policy impact camera / mic / geo not explicitly denied fix Permissions-Policy: camera=(), microphone=(), geolocation=()
Document language, image alts, form labels, heading outline from the fetched HTML. Enough to catch launch-blocking gaps — not a full WCAG audit by a human.
LOW <html> missing lang target https://northstar.app evidence root element has no lang · 6 images without alt impact screen readers guess the language fix <html lang="en">
CDN, WAF, CMS, server tokens, JS libraries. Used to pick the right fix snippets (Nginx vs Cloudflare vs WordPress) and to drive the hunt playbook toward paying checks for that stack.
target northstar.app CDN Fastly WAF — CMS Ghost Server nginx/1.25 playbook check /ghost/ /rss/ /.well-known/security.txt
2.1R is a big model with a small mandate: observe, rank, explain. It will not write exploit PoCs, run authenticated attacks, or scan past program scope.
What 2.1R is — and what it is not.
RX VeX 2.1R is the current reasoning revision of the RX VeX core — a large cyber language model with constrained decoding and live-fetch grounding. R is the reasoning line.
No. It fetches the URL you authorize, then expands only to in-scope hosts, historical URLs, and signatures. Out-of-scope names are tagged, not attacked.
No. 2.1R maps and ranks. Paste the Hunt graph into Burp or Caido if you need an interactive proof. The report stays existence-and-fingerprint so the product can be hosted.
A /100 score, seven category bars, ranked findings with evidence, a fix roadmap with snippets, HackerOne-ready markdown, PDF export, and rescan diffs.
Sign in, paste an authorized URL, and let 2.1R score all seven lanes.