onboarding page
Dashboard page
Payment form
Payment approval
Activity
Partner inbox
Architecture

Vouchsafe — Project Story

Inspiration

Addiction rarely fails in a dramatic moment — it fails in the thirty seconds right after money lands in someone's hands. A paycheck arrives, and before any real decision gets made, it's already gone: one bet, one bar tab, one impulsive transfer. We watched this pattern play out for people close to us, and the tools that exist today only offer two extremes — total, rigid blocking (which people abandon the first time it's inconvenient) or nothing at all.

We also noticed a second, quieter problem: money sent for a real purpose — tithe, rent, supporting family — often arrives with zero context, forcing an awkward phone call just to explain what it was for. We realized both problems shared the same missing piece: purpose. If every payment carried its reason, two things become possible at once — no more explaining yourself over the phone, and a trusted person can make a real judgment call on the transactions that actually need one.

That's Vouchsafe: not a lock, not a lecture — a second set of eyes, exactly when it matters.

What it does

Vouchsafe adds a human checkpoint to spending:

Every transaction requires a purpose — a category (Rent, Chama, Alcohol, Betting, etc.) and an optional note, so recipients and the sender's own records always have context, no phone call required.
Risky transactions require approval — anything above an agreed threshold, or tagged under a high-risk category, is held as pending until a trusted accountability partner approves or declines it.
Partners see only what they need to — amount, category, and recipient. Never the sender's balance or incoming money. Full financial oversight isn't the goal; a second opinion at the right moment is.
The category engine catches disguised spending — not just large transactions. If "Rent" gets used twice in a month, or someone starts routing the same recipient through three different category labels to dodge scrutiny, Vouchsafe flags it with clear context for the partner to review.

How we built it

Backend: FastAPI, chosen for speed of development and automatic API docs — useful both for our own testing and for demoing live to judges.
Database: SQLite, kept deliberately simple for the hackathon timeline — zero setup, one file, no server configuration eating into build time.
Payment layer: a mocked/simulated payment service standing in for a real processor (M-Pesa was our original inspiration and target market). It logs and returns realistic responses so the full approval-to-execution flow works end-to-end without real payment integration overhead.
Category validation engine: the core differentiator — rules for frequency limits, duplicate-use detection, always-flagged categories, and amount-threshold checks, all running server-side before any transaction is allowed through.
Frontend: a lightweight React interface with two views — the sender's transaction flow and the partner's approval queue — so the full loop is visible and demoable in one sitting.

Challenges we ran into

Scoping down without losing the point. The full product vision includes real payment processor integration, mobile apps, SMS notifications, and much richer pattern detection. Fitting the core mechanic — purpose tagging plus category-aware approval — into a 14-hour build meant cutting hard and being honest in our writeup about what's simulated versus production-ready.
Making the category rules genuinely useful, not just decorative. It would have been easy to build a simple amount-threshold check and call it done. Getting the duplicate-detection and always-flagged logic actually working — and explainable to the partner in plain language — took real iteration.
Designing partner permissions correctly. Deciding exactly what a partner can and can't see took real thought: too little visibility and their approval is meaningless; too much and the tool becomes financial surveillance instead of support.

Accomplishments that we're proud of

A working end-to-end flow: a transaction gets created, validated against category rules, routed to a partner when it should be, and resolved — all functioning together, not just as separate pieces.
Category-abuse detection that goes beyond a simple dollar-amount check, catching the exact evasion pattern (splitting or relabeling spending) that a naive threshold system would miss entirely.
A permissions model that respects both the sender's dignity and the partner's need for real context — no full financial surveillance, just the information necessary to make one good decision.

What we learned

How real financial compliance systems detect structuring (splitting transactions to evade oversight) — and how directly that same logic applies to personal accountability tools, not just banks.
The importance of designing friction thoughtfully: too much, and people abandon the tool the first time it's inconvenient; too little, and it doesn't protect anyone. Every rule we built, we tried to justify against that balance.
That a good accountability system isn't about restriction — it's about giving someone the context to make one good decision, at one important moment.

What's next for Vouchsafe

Real payment processor integration, starting with M-Pesa's Daraja API — the market this was originally designed for, where mobile money is the primary way people transact.
Mobile-first apps for both the sender and the accountability partner, with push notifications replacing our current in-app/simulated alerts.
Backup accountability partners and smarter fallback handling for when a partner is unreachable, so the tool never becomes a hard blocker on someone's daily life.
Richer structuring detection, tuned with real usage data rather than hardcoded thresholds.
Piloting with real users and their real accountability partners — the actual test of whether this helps people, beyond a hackathon demo.