Gauge

Inspiration

Your digital assets already have great dashboards. Open any brokerage or crypto app and you see what you paid, what it's worth right now, whether you're up or down, and a chart of how it changed. Your physical assets get none of that. For most people, a car, a home, a watch, or a card collection is worth more than their brokerage account, yet nobody can tell them what those things are worth today. That gap hurts when you sell something, file an insurance claim, or just want to know your real net worth. We wanted to give the things you can touch the same treatment as the things on a screen.

What it does

Gauge tracks what everything you own is worth.

Add anything, one question at a time. Pick a category (vehicles, real estate, watches, jewelry and metals, collectibles, electronics, home goods, tools) and Gus, our boiler-inspector mascot, asks only the questions that category needs.
Real market values, not guesses.
- Cars: paste a VIN and it decodes the year, make, and model from NHTSA, then pulls retail and trade-in values from VinAudit. Cars can refresh their value every month automatically.
- Homes: an address becomes an estimate from RentCast, based on comparable homes nearby, with a likely range.
- Trading cards: live market prices from TCGdex and Scryfall.
- Gold, silver, platinum: melt value from the live spot price.
- Everything else: an AI price check (Gemini) gives a ballpark, clearly labeled as a rough estimate so you always know how solid a number is.
A portfolio view for physical things. Your total value, cost basis versus today's value, gains and losses, a breakdown by category, and a value history that grows as values update.
Liquidity, not just price. Physical assets are hard to sell fast, so every item tracks both fair market value and quick-sale value.
Snap it from your phone. Scan a QR code and photos taken on your phone go straight to your desktop session, where AI helps identify the item.
Insurance-ready. Export your whole inventory as a PDF.

How we built it

Frontend: React 19, TypeScript, Vite, Tailwind CSS v4, and shadcn/ui, with React Router, TanStack Query, React Hook Form + Zod for validation, and Recharts for charts. PDFs are generated with @react-pdf/renderer.
Backend: an Express 5 API in TypeScript with Mongoose on MongoDB Atlas. The same Express app runs locally and as a Vercel serverless function, with a Vercel cron job for the monthly vehicle value updates.
Auth: Clerk. Every database query is filtered by the signed-in owner, so one user can never read another user's things.
Data model: money is stored in integer cents. Each value is a dated observation (fair market, quick-sale, replacement cost), so history is kept instead of overwritten. The portfolio total on any date is the sum of each asset's latest observation up to that date:

$$V(t) = \sum_{a \in \text{assets}} v_a\big(\max{\, s \le t : a \text{ has a value on } s \,}\big)$$

Metals math: melt value comes straight from the spot price per troy ounce:

$$\text{melt} = \frac{\text{weight (g)}}{31.1035} \times \text{purity} \times \text{spot price}$$

So a 10 g, 12k (50% gold) necklace at $\$3{,}997$/oz melts to about $\$643$.

Challenges we ran into

Honest numbers are hard. There's no free source of historical home or car values, so instead of inventing a fake history we record real dated observations and let the chart grow over time. We also label AI estimates as rough and explain that melt value is a floor for jewelry, not a sale price.
APIs that didn't cooperate. Our watch pricing API never returned usable prices, so we rebuilt watch pricing on top of the AI price check fallback.
Tiny free quotas. RentCast's free plan allows 50 lookups a month. We added caching on both the browser and the server, plus per-user rate limits, so repeated clicks don't burn through quota.
Privacy. VINs and home addresses are sensitive, so lookups are POST requests (never in URLs or logs), and an address is only sent to RentCast when you ask for an estimate.
Three people, one codebase, one weekend. Splitting the UI and the API work, then merging it all without breaking each other's features, took real coordination.

Accomplishments that we're proud of

Seven live data sources behind one simple "what is it worth?" flow.
A step-by-step add flow that feels friendly instead of like filling out a tax form.
Every number on screen says where it came from.
Deployed end to end with real sign-in, a real database, and a scheduled job.

What we learned

How valuation actually works for different asset classes: comparables for homes, mileage and condition adjustments for cars, spot price and purity for metals.
Designing around unreliable third-party APIs: fallbacks, caching, rate limits, and clear messages when something isn't available.
Running an Express API as a serverless function, with scheduled jobs.

What's next for Gauge

Monthly automatic updates for every category, not just cars, so every chart moves over time.
Value alerts ("your car dropped 8% this quarter").
Receipt and warranty storage alongside each item.
Sharing a read-only inventory with an insurer or a family member.