The brief
Scams arrive on WhatsApp as links, screenshots and voice notes, and the people most at risk will never install a security app. The checker has to live where the messages already are and answer in seconds with a verdict a grandparent can act on.
What we built
- Link checks that unwind redirects, look at domain age and detect look-alikes of Nigerian brands.
- Verdicts from an AI model under fixed rules the model can only make stricter, never more lenient.
- Screenshot, PDF and voice-note understanding.
- Credits and checkout through Paystack in naira or US dollars, plus a Family Pass.
- Group scanning for WhatsApp admins, a reviewer console, disputes and data deletion on request.
- A public scam radar page and a partner threat-feed API.
Where it stands
Phases 1 to 5 are built with a test suite, smoke tests and an evaluation harness. CI and deployment files are ready; the name and launch date are not final.
Stack
- Node 22
- Fastify
- Prisma
- PostgreSQL
- pgvector
- Redis
- BullMQ
- WhatsApp Cloud API
- Telegram Bot API
- Claude
- Google Web Risk
- Paystack