Scanara

I built Scanara because I kept seeing the same pattern: teams shipping AI features have no idea whether they're "high-risk" under the EU AI Act, and even when they do, the required documentation (Annex IV — 9 sections) takes 3-6 months to produce by hand. Every EU AI Act compliance tool I could find treats this as a legal/governance problem: questionnaires, dashboards, consultants. Scanara treats it as an engineering problem. It's a scanner that reads your actual code and documents, classifies your AI systems by risk tier (Annex III), and generates the compliance dossier from what it finds — not from what you tell it. What it checks: 44 code-detectable articles (missing human-oversight hooks, no audit logging, no input validation on model-facing endpoints, etc.) + 22 document-level obligations, across 8 EU languages. Merge gate blocks non-compliant PRs if you want it to (fail-closed, not fail-open — a lot of compliance middleware silently allows on timeout, which is worse than no check at all).