BoundBuy AI – Project Description
Inspiration
Traditional procurement is often complicated by scattered supplier information, changing prices, budget overruns, hidden risks, and a lack of transparency. Buyers frequently have to manually compare products, verify availability, calculate costs, and manage payment decisions.
We wanted to make purchasing more intelligent without taking control away from the buyer.
This inspired us to build BoundBuy AI around one simple principle:
"You set the buying rules. AI finds the options. You approve. PayPal pays."
Our goal is to combine AI-powered procurement intelligence with secure digital payments, giving buyers greater control, transparency, and confidence throughout the purchasing journey.
What it does
BoundBuy AI is an AI-powered procurement copilot that transforms natural-language purchasing requests into structured, policy-controlled purchasing decisions.
Its key capabilities include:
- AI Purchase Planner: Understands purchasing requirements, extracts quantities, budgets, deadlines, and restrictions, and generates structured purchase plans.
- Smart Product Catalog: Enables users to explore products, compare prices, check availability, and review supplier information.
- Procurement Intelligence: Provides purchase insights, what-if price simulations, budget analysis, and explainable risk reviews.
- Purchase Pact: Creates an approved purchasing snapshot with budget limits, supplier conditions, and cryptographic integrity verification. Unauthorized changes invalidate the approval.
- PayPal Integration: Uses PayPal Orders v2 Sandbox for payment order creation and buyer-approval verification.
- Orders and Digital Receipts: Provides order history, payment references, and downloadable PDF receipts.
- Shipment Tracking: Integrates with AfterShip to support shipment tracking and delivery updates.
- AI Assistant: Offers contextual answers about purchases, orders, payment information, and delivery details.
BoundBuy AI keeps the buyer in control while using AI to simplify complex procurement decisions.
How we built it
We built BoundBuy AI using a modern full-stack architecture.
Frontend:
- React 19
- TypeScript
- Vite
- Tailwind CSS
- Recharts
- Lucide Icons
Backend:
- Node.js
- Express.js
- REST APIs
AI Integration:
- Google Gemini API for natural-language purchase planning, procurement insights, and contextual assistance.
- Deterministic fallback engine for reliable request processing during AI service outages or rate limits.
Payments:
- PayPal Orders v2 REST API
- PayPal Sandbox authentication and checkout
- Server-authoritative pricing and payment validation
Shipment Tracking:
- AfterShip Tracking API
- Separate read and write API credentials
- Webhook signature verification
Security and Integrity:
- SHA-256 Purchase Pact snapshot hashing
- Server-side budget and price validation
- Approval invalidation when purchasing conditions change
- Zod-based request validation
- HMAC-SHA256 webhook verification
The application also includes persistent local storage with an architecture that supports MongoDB Atlas configuration.
Challenges we ran into
One of our biggest challenges was making AI-generated purchasing recommendations reliable. Natural-language requests can contain incomplete information, ambiguous quantities, and changing budget constraints. We addressed this with structured extraction, validation, and a deterministic fallback engine.
Integrating PayPal also required careful handling of its payment lifecycle. We had to ensure that an order could not be captured before buyer approval and that payment amounts were independently validated on the server.
Another major challenge was preventing changes to approved purchasing conditions. We implemented cryptographic Purchase Pact snapshots so that changes to prices or quantities invalidate previous approvals.
We also worked on handling external API failures, tracking authentication, webhook security, delivery scheduling, and maintaining consistent order and payment records.
These challenges pushed us to focus not only on building features, but also on making the entire purchasing workflow dependable.
Accomplishments that we're proud of
- Built a complete AI-powered procurement experience connecting purchase planning, product discovery, risk analysis, approval, payment, and tracking.
- Developed a Purchase Pact mechanism using SHA-256 hashing to detect unauthorized changes to approved purchasing conditions.
- Integrated PayPal Orders v2 Sandbox with server-side order creation and buyer-approval validation.
- Implemented explainable risk reviews and what-if purchasing simulations.
- Created a deterministic AI fallback mechanism that avoids inventing product IDs, prices, or inventory.
- Integrated AfterShip API authentication and secure webhook verification.
- Developed downloadable PDF receipts and order-history functionality.
- Completed a comprehensive technical audit with 124 automated tests passed and zero failed in the reported test run.
Our biggest accomplishment is bringing procurement intelligence, buyer-controlled approvals, cryptographic safeguards, and digital payments together in one platform.
What we learned
We learned that building a useful AI-powered application requires much more than connecting an AI model to a user interface.
AI recommendations must be validated against real product data, budgets, supplier conditions, and business rules.
We also learned the importance of server-side payment verification, idempotency, cryptographic integrity, secure API integrations, reliable fallback mechanisms, and clear separation between payment status and delivery status.
Most importantly, we learned that AI should assist people in making better decisions rather than silently making financial decisions on their behalf.
What's next for BoundBuy AI
We envision expanding BoundBuy AI into a more comprehensive procurement platform.
Our next priorities include:
- Completing and verifying the full buyer-approved PayPal Sandbox capture lifecycle.
- Enabling MongoDB Atlas for scalable remote persistence.
- Expanding supplier onboarding and multi-supplier procurement workflows.
- Improving Gemini-powered price intelligence and supplier recommendations.
- Introducing more advanced fraud detection and procurement anomaly analysis.
- Expanding real-time shipment tracking and verified delivery updates.
- Supporting additional payment providers and regional currencies.
- Developing enterprise procurement dashboards and team-based approval workflows.
Our long-term vision is to make procurement smarter, safer, and more transparent through AI-powered intelligence combined with buyer-controlled financial decisions.
Built With
- api
- artificial
- css
- express.js
- gemini
- hmac-sha256
- icons
- intelligence
- javascript
- lucide
- node.js
- orders
- paypal
- react
- recharts
- rest
- sandbox
- sha-256
- tailwind
- typescript
- vite
- zod