Kanakku
Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

Some of my friends lost money last year. It wasn't bad luck: it was a broker's ad promising "sure-shot" returns, a WhatsApp tip, a policy sold as "your money doubles." I was luckier. My father has invested in Indian markets for nearly 25 years, and whenever an offer looks good, I can ask him first. My friends couldn't. Neither can most families. Complaints of unfair selling against life insurers alone rose to 26,667 in FY25 (IRDAI). Kanakku gives everyone that person to ask. Hold one button and say, "This man says I pay ₹50,000 a year and get ₹15 lakh after 15 years. Is it good?" Then show the brochure. Alfii, the curator, calls only the specialists the question needs: Sankz for banking and loans, Dani for funds and insurance, Vick for stock tips. They argue the case in their own voices and answer red, yellow or green, with the exact questions to ask the seller. A family member sees the alert in the app and can leave a note. When someone else is in the room, Kanakku keeps your own numbers off the speaker. Trust is the whole product. The AI never types a number. It writes references that code fills in from deterministic calculations, and any argument without a source is thrown out before the vote. We tested this on Meta's API across 29 financial scenarios: no risky offer came out green, and an independent recalculation of 22 figures found 0 mismatches. Kanakku explains offers; it never tells you which stock to buy. Next: with the user's consent, connect bank, insurance and investment data through India's Account Aggregator framework, so Alfii sees the whole family picture in one place. The more a family's details Kanakku has, the better its answers get.