Where's the Money

A mobile-first financial copilot that turns transactions into explainable decisions, cash-flow radar and actionable plans.

Comment

Inspiration

Most personal finance apps tell users what already happened. Where's the Money is designed to answer a more useful question: what does this mean for the decisions I need to make next?

The project turns financial history into an understandable, auditable picture of cash flow and then helps the user plan around goals, recurring expenses and future choices.

What it does

Where's the Money is a mobile-first financial assistant that combines deterministic financial logic with an AI explanation layer.

It can:

  • import transactions from OFX, CSV, TXT, XLS and XLSX files;
  • connect to Open Finance through Pluggy;
  • classify common transactions with deterministic rules;
  • surface ambiguous transactions for human review instead of guessing;
  • learn reusable user-defined categorization rules;
  • detect weekly, biweekly and monthly recurring transactions;
  • show a Cash Radar with a 7-day Free view and a 30-day Pro view;
  • separate upcoming financial pressure into a dedicated Watch Out / Fique de olho experience;
  • analyze Essentials / Flexible / Future spending with editable budgeting references;
  • support multiple financial goals in a conversational planner;
  • calculate deterministic allocation scenarios, compound-interest simulations and financing scenarios;
  • use Gemini with Google Search grounding for contextual market research while keeping financial calculations outside the AI model;
  • offer Free and Pro access through native RevenueCat entitlements.

How we built it

The product uses a React + TypeScript web core deployed through Cloudflare Workers and Static Assets. The native mobile layer lives in Expo and loads the product through a WebView while providing the native integrations required by the hackathon.

RevenueCat is the source of truth for the real pro entitlement, including Paywall, restore purchases and Customer Center. The WebView communicates with the native RevenueCat layer through a typed bridge rather than duplicating the product UI.

The architecture deliberately separates evidence and deterministic calculations from AI. Rules, transaction direction, recurrence, budget capacity and scenario mathematics are computed locally and reproducibly. Gemini is used only where natural-language interpretation, explanation or external context adds value.

Challenges

A major challenge was converting an already substantial web financial engine into a credible native mobile experience without duplicating logic or weakening the product architecture.

Other challenges included:

  • keeping RevenueCat as the only source of truth for paid entitlement;
  • ensuring demo access never impersonates a real purchase;
  • connecting native purchase state safely to the WebView UI;
  • designing AI features that cannot silently alter financial truth;
  • handling missing API keys and integrations with safe fallbacks;
  • preserving a useful Free product while making the Pro upgrade meaningful.

Accomplishments we're proud of

  • A deterministic financial engine that remains useful without AI.
  • A reproducible synthetic Demo Pro mode for judges that does not overwrite real user data or create fake RevenueCat entitlement.
  • Native RevenueCat integration with pro entitlement, Paywall, restore and Customer Center.
  • A typed WebView bridge between Expo and the product UI.
  • Open Finance integration through Pluggy Sandbox.
  • A Cash Radar that distinguishes known data from uncertainty instead of overstating predictions.
  • A conversational planner where facts only become structured plan data after user confirmation.
  • A public build-in-public series documenting real progress, failures and design decisions during Shipaton.

What we learned

Building in public made technical failures part of the product feedback loop instead of something to hide. Runtime problems, integration gaps and UX trade-offs became visible checkpoints that could be explained, corrected and used to improve the next iteration.

The project also reinforced an important architecture principle for financial software: AI can explain, research and help structure intent, but balances, recurring patterns, budget capacity and scenario math should remain auditable and deterministic.

RevenueCat

Where's the Money uses RevenueCat as the native source of truth for the pro entitlement.

The Free tier includes transaction import, Inbox/manual review, deterministic classification, up to three active rules, a 7-day Radar preview and basic goals/planning.

Pro adds automatic Open Finance, a full 30-day Radar, unlimited rules, advanced recurrence and pressure details, the full conversational planner, contextual research and advanced scenarios/simulations.

The native Expo app provides RevenueCat Paywall, restore purchases and Customer Center. The browser build remains Free unless the user explicitly opens the synthetic Demo Pro experience.

Build in Public

The entire Shipaton journey is being documented publicly as a video series. The content shows development progress rather than only the final polished app: code, mobile migration, failed builds, runtime issues, design decisions, RevenueCat integration and the work required to turn the initial web product into a real mobile submission.

What's next

Before final submission we are focusing on native Android validation, final RevenueCat configuration, production-level QA, final UI polish, submission screenshots/icon, the public demo video and the final Devpost submission.

Built With

Share this project:

Updates