MandateLens

## Inspiration

We wanted to grow in the AI space and build our first AI project together. We also wanted to choose a problem that could be useful to merchants and PayPal customers. That led us to explore how an AI shopping assistant could help people shop while keeping them informed and in control.

## What it does

Mandate Lens demonstrates an agent-assisted shopping journey through PacePoint, a running-shoe storefront. The shopper sets a budget, reviews and approves a product selection, and asks the Guard to check the cart. The Guard evaluates the cart against the demo mandate and merchant catalog, then records its decision and reasons. After a final ALLOW, the shopper approves checkout through PayPal Sandbox. AgentPulse displays the completed journey.

The AI client can open the shop and retrieve a journey summary through MCP. The shopper approves the selection, Guard review, and payment in the interface.

## How we built it

We built the APIs with Java and Spring Boot, the PacePoint storefront and AgentPulse dashboard with React and Vite, and the MCP server with Python and FastMCP. PostgreSQL stores the catalog, carts, decisions, and journey events, with Flyway managing database migrations.

Merchant owns the catalog, carts, and PayPal integration. Mandate Lens runs the deterministic Guard and records journey decisions and events. We deployed the services on Render and used Postman to test the APIs. The PayPal Sandbox integration uses the Orders API v2.

## Challenges we ran into

One of our biggest challenges was finding a problem statement that could provide value to PayPal’s customers and to PayPal overall. Once we chose agent-assisted shopping, we also had to connect several services while keeping the shopper’s approvals, Guard decisions, PayPal checkout, and journey records in sync.

## Accomplishments that we’re proud of

We completed the project while balancing our full-time jobs. We’re proud to have built a complete, interactive demo and developed an idea that we believe could be useful to merchants and PayPal: making agent-assisted purchases easier to understand and keeping shoppers involved in important decisions.

## What we learned

We learned a lot about AI, how to think through a system as a whole, and how to integrate AI so systems can become more intelligent and automated. We also saw how those capabilities can help people, and why clear approval steps and decision records matter when AI is involved in a purchase.

## What’s next for Project Emeralds

As Project Emeralds, we want to keep finding meaningful problems and exploring how AI can help automate them. For Mandate Lens, we’d love to explore whether this idea could be taken further with PayPal and potentially developed into a production solution.