ShopperRelay

Shopping help from someone you trust, with rules made to fit you. AI understands. Rules enforce. People approve. PayPal executes.

Comment

Inspiration

Sarah helps everyone. ShopperRelay helps Sarah.

Sarah is a busy working woman. Her mom needs help making safe purchases. Her son needs things for school. Her husband needs a new jacket for an important meeting. And Sarah? She has her own shopping to do.

We built ShopperRelay because helping the people you love shouldn't mean doing everything yourself.

We wanted to make shopping assistance easier, whether it comes from family, a trusted advisor, or an AI agent, without requiring anyone to surrender control over their money.

What it does

One platform. Four ways to help.

  • Mom shops independently. ShopperRelay's AI discovers a hidden $39 monthly subscription on a $149 blood pressure monitor. Her rules bring Sarah in, and PayPal only holds the money: nothing is charged unless Sarah approves. If someone pressures Mom to loosen a rule, the change waits 24 hours and Sarah is told.

  • Leo asks, Dad decides. Sarah's rules route electronics to Marcus, who reviews the product next to Leo's budget. Leo never needs his parents' payment credentials.

  • Marcus gets trusted advice. Priya, his paid shopping assistant, finds the right jacket. Marcus approves this one jacket up to a limit he sets; PayPal charges only the real price and releases the rest. Priya invoices for her time through PayPal.

  • Sarah gets help, too. Her AI shopper finds a Secret Santa gift under $25 and buys it on its own with her saved PayPal, only within her rules. She still shops for herself as well, and every charge to her PayPal, whether hers, her AI shopper's, or one she approved for Leo, lands in one table she can sort, search and download.

ShopperRelay makes assisted shopping possible without giving anyone unlimited purchasing authority.

How we built it

We combined AI, enforceable purchasing rules, real product information, and PayPal payments into one shopping experience.

Google Gemini understands purchasing instructions written in everyday language and reads checkout screenshots to identify hidden conditions. A separate rules engine, with no AI in it, enforces spending permissions. PayPal handles payment authorization, capture, release, saved payment permissions through Vault, and invoicing. Before any capture, the amount PayPal is actually holding is read back and checked against the rules again.

Our AI shopper independently plans searches and selects products, but code verifies each listing against the user's rules before any purchase. PayPal's authorization provides an additional spending ceiling.

Channel3 supplies real product listings, prices, retailers, categories, and product grouping. AG Grid powers searchable, sortable shopping and payment history with CSV export. Bryntum provides an occasions calendar for scheduling AI-assisted purchases.

We built the application with Next.js, React, TypeScript, Vercel, and Neon Postgres, supported by 174 unit tests and end-to-end tests that run every flow against the PayPal sandbox. Retailer order fulfillment is simulated.

Challenges we ran into

Our biggest challenge was making AI useful without giving it unrestricted access to someone's money. We solved that by separating intelligence from authority: AI recommends, code enforces, and people remain in control.

We also had to balance protection with independence. Shopping rules should help older adults make informed decisions without taking away their ability to make their own choices.

Finally, we needed shopping permissions to be specific and enforceable, ensuring a one-time purchase approval couldn't become unlimited spending authority.

Accomplishments that we're proud of

We're proud of building a working prototype that demonstrates four distinct shopping relationships through one shared system.

We integrated real PayPal sandbox payment flows, AI-assisted checkout analysis, purchasing rules, approval routing, controlled AI shopping, and professional invoicing.

We also built an interactive demo that anyone can explore without registering, with a guided tour inside the demo and a step-by-step test script for every claim in our video.

What makes us proudest is that every feature serves a real human need, not simply a technical demonstration.

What we learned

The future of AI shopping isn't just about finding the right product. It's about knowing who has permission to buy it.

We learned that AI can make purchasing decisions easier, but financial control must remain predictable, transparent, and enforceable.

We also discovered how one permission-based shopping system can support families, caregivers, professional shopping advisors, and autonomous AI agents.

What's next for ShopperRelay

Our next steps are connecting real merchant checkout systems, adding direct notifications, expanding shopping relationships, and improving accessibility for families and older adults.

We envision ShopperRelay becoming a trusted way to shop with help, whether that help comes from someone you love, a professional you trust, or an AI agent.

Try the live demo: no account required, PayPal sandbox only. To pay, choose Debit or Credit Card and use the test card 4111 1111 1111 1111, any future date and any CVV.

Follow the test script: see and test every workflow from our video.

AI understands. Rules enforce. People approve. PayPal executes.

Sarah helps everyone. ShopperRelay helps Sarah.

Built With

  • ag-grid
  • agentic-commerce
  • ai-agents
  • bryntum
  • channel3
  • gemini-api
  • google-gemini
  • invoicing
  • neon
  • next.js
  • node.js
  • paypal
  • paypal-javascript-sdk
  • paypal-orders-api
  • paypal-sandbox
  • paypal-vault
  • paypal-webhooks
  • postgresql
  • react
  • remotion
  • tailwindcss
  • typescript
  • vercel
  • vitest
  • zod
Share this project:

Updates

Submission history