Product
A crawler, an audit engine and a reporting layer that agree with each other
Most teams stitch together a headers checker, a Lighthouse run, a spreadsheet and a slide deck. WebAuditOS is those four things as one system, with one score model and one evidence trail.
How a scan runs, in four stages
The crawler
You give a root URL, a depth and a focus mode. The crawler follows internal links, obeys robots.txt before it fetches rather than after, records every response and builds a page inventory with type classification, status codes, canonicals and indexability.
Crawl budget is spent breadth-first: pages are taken in discovery order and the run stops at your budget. Admission happens before fetching rather than after, so the budget is a hard promise about how many requests reach the site - not a number we trim to afterwards. The crawl does not rank routes, so it goes wide before it goes deep.
- Depth 1–6, with per-plan page ceilings
- Focus modes: security, SEO, full audit, client audit
- Automatic page-type classification into 16 categories
- Sitemap ingestion when one is published
- Redirect chain and loop resolution
The audit engine
Every discovered page runs through the check library. Checks are grouped into nine weighted categories and each one produces a finding with a severity, a status, evidence and a detection method.
Detection method is the part most tools skip. Auto-detected means we observed it directly. Inferred means we derived it from strong public signals and attached a confidence value. Manual means it is server-side state that no external scan can honestly claim to know.
The scoring model
Category scores start at 100 and lose points per finding: the severity, scaled by how confirmed the finding is and by how much of the site it affects. The overall score is those nine category scores weighted and averaged. Which page a problem is on does not change its cost - only how many pages it is on.
Not-applicable items carry no penalty at all, so a site without file uploads is neither rewarded nor punished for a control it doesn't need. Manual items are the same: held out of the score entirely and reported as unresolved rather than silently failing or silently passing. /methodology publishes every number, rendered from the module the engine computes with.
- Overall and per-category scores from 0–100
- Letter grade for executive summaries
- Confidence value per score
- Explained deltas between any two scans
The reporting layer
The same findings feed five report presets - full audit, security posture, SEO and content, accessibility and UX, and executive summary. A preset is a selection of category sections, and the checkboxes show exactly what it selected, so you can add or remove any of them. Every paid plan can export a print-optimised PDF and publish a read-only link that expires; Agency and Enterprise can also rebrand the output.
Reports track movement, not just state: fixed versus still-open, score change by category, and outstanding manual attestations that need a human signature.
Scope
What runs today, and what does not exist yet
A scan is a real crawl. The engine fetches pages over the network, renders a sampled subset in Chromium, runs the check library against what it observed, and computes the score from those findings - there is no simulated path in a signed-in workspace. Demo mode is a separate, clearly-labelled fixture layer for reviewing the interface without a backend.
- Running: crawler, Chromium rendering and screenshots, the check library, scoring, scheduled monitoring, alerts, Stripe billing, PDF export, e-mail for invitations and resets
- Built but not proven end to end: nothing in this product has been walked from sign-up to shared report by a human, which is the largest gap we know about
- Not built: outbound webhooks, chat and issue-tracker connectors, SSO/SAML, SCIM, per-workspace data residency, vertical checklist templates, scheduled e-mail digests to a client, a customer-supplied report domain
- The operating company is published on /contact and in the footer (ADMI ENTERPRISES s.r.o., IČO 29581010); the VAT position is stated there too. There is still no staffed support channel - [email protected] reaches one person
Run the four stages on your own domain
Free, no card. We read only public pages and respect robots.txt.