Real-time, fully explainable fraud decisioning: 34 signals across seven categories, seven named rule packs, and ALLOW / REVIEW / BLOCK outcomes that always carry their evidence.
Outputs on this page were captured by running the repository's own Python engine over ten synthetic events on 19 November 2026. The browser replays those decisions — no model runs in the page, and no live data is involved.