-
-
PayGuard AI's working shopping dashboard with product catalogue, AI assistant and secure cart.
-
PayGuard AI — intelligent shopping recommendations with human-controlled PayPal payments.
-
Responsive full-stack shopping application with curated products, AI assistance and cart management.
-
Natural-language shopping assistance recommends three products totaling $55.97.
-
BudgetLock validates a $55.97 cart against a $100 budget and requires explicit customer approval.
-
PayPal Sandbox checkout enables simulated buyer authorization for a $55.97 USD purchase
-
Successful PayPal sandbox capture: COMPLETED, $55.97 USD, verified by the backend.
PayGuard AI — Shopping Intelligence. Human Control.
Inspiration
AI is transforming online shopping, but letting AI make purchasing decisions introduces an important problem: how do we ensure that users remain in control of their money?
AI assistants can recommend products, but their suggestions may exceed a customer's budget, include inaccurate prices, or lead to unwanted purchases.
We built PayGuard AI to demonstrate a safer model for AI-assisted commerce: intelligent product discovery combined with independent budget protection, transparent pricing, explicit human consent, and verified PayPal payments.
Our guiding principle is simple: AI can recommend, but humans decide when to pay.
What it does
PayGuard AI is a full-stack, consent-first shopping assistant integrating Google Gemini with PayPal's developer platform.
Its primary features include:
- AI-powered shopping assistant
Customers describe what they need in natural language. The assistant recommends relevant products from a curated catalogue, considering the requested spending limit.
- BudgetLock protection
A server-side validation engine independently calculates product prices, checks quantities, and prevents orders from exceeding the customer's budget. AI-generated recommendations cannot override these rules.
- Human-controlled checkout
Before checkout, customers review their selected products, verify the total, and explicitly approve initiating a PayPal payment.
- PayPal Sandbox integration
PayGuard uses PayPal's sandbox environment to create checkout orders, obtain buyer approval, capture payments, and verify completed transactions.
- Resilient AI assistance
The application supports configurable Gemini models and fallback behavior. When the external AI service is unavailable, local catalogue assistance remains available without pretending that an AI request succeeded.
- Secure payment verification
The backend validates payment status, currency, amount, and transaction information before displaying a verified payment receipt.
How we built it
We developed PayGuard AI as a browser-based full-stack application.
Frontend: HTML, CSS, and JavaScript provide a responsive interface with a product catalogue, conversational shopping assistant, spending-limit controls, shopping cart, checkout approval, and payment receipt.
Backend: Node.js handles API requests, product validation, shopping recommendations, budget enforcement, payment orchestration, and error handling.
Artificial Intelligence: Google Gemini API processes natural-language shopping requests. The application validates AI recommendations against trusted catalogue items and server-controlled prices.
Payments: PayPal REST APIs and the PayPal JavaScript SDK support sandbox order creation, customer authorization, payment capture, and transaction verification.
Security: Sensitive API credentials remain on the backend. Additional protections include independent price validation, spending limits, consent requirements, request validation, origin checks, and duplicate-capture safeguards.
Testing: We implemented automated tests covering recommendations, payment validation, security controls, fallback behavior, and other important application functions.
Challenges we ran into
One major challenge was handling AI service delays and temporary unavailability. We implemented timeout controls, fallback behavior, and server-side validation so that shopping assistance remained useful without compromising payment security.
Another challenge was integrating PayPal sandbox checkout with test merchant and buyer accounts. We encountered sandbox checkout errors, investigated account configurations, and validated the checkout process using an appropriate sandbox setup.
We also needed to ensure that AI recommendations could not manipulate product prices or bypass customer spending limits.
Accomplishments that we're proud of
Built a functioning AI-assisted shopping application with frontend and backend integration.
Integrated the PayPal developer platform into the central checkout workflow.
Successfully completed a $55.97 USD PayPal sandbox payment, with a completed capture status and transaction identifiers.
Demonstrated a shopping cart that remained within a $100 customer budget.
Implemented explicit customer consent before initiating checkout.
Achieved 24 passing automated tests with zero failures.
Published the project source code and documentation in a public GitHub repository.
What we learned
This project strengthened our understanding of full-stack development, AI integration, REST APIs, payment authentication, sandbox testing, error handling, and transaction verification.
Most importantly, we learned that responsible agentic commerce requires a clear separation between AI recommendations and financial authorization.
An AI assistant should help users make decisions without taking away their control.
What's next for PayGuard AI
We plan to explore:
More advanced AI-powered product comparisons and shopping assistance.
Improved personalization and customer preferences.
Merchant integrations beyond the demonstration catalogue.
More extensive automated security and payment testing.
Customer accounts, order history, and enhanced transaction monitoring.
Production-grade deployment and stronger account-level security controls.
Our long-term vision is to make AI-assisted commerce more intelligent, transparent, and trustworthy.
PayGuard AI: Let AI handle the recommendations. Keep humans in control of payments.
GitHub: https://github.com/prabhudev30/PayGuard-AI-PayPal-Hackathon-2026
Environment: Local web application using PayPal Sandbox. All demonstration payments use simulated funds.