The Problem

Managing refunds, returns, and warranty claims can be confusing. Customers must keep track of purchase information, return deadlines, retailer policies, supporting evidence, and follow-up communications. Missing a deadline or forgetting an important document can make the process even more frustrating.

Our Solution

Refund Guardian AI is a refund-management assistant that helps users organize their purchases and manage the steps involved in pursuing a refund, return, or warranty claim.

The application provides a centralized dashboard for managing cases, tracking deadlines, analyzing potential eligibility based on user-provided policy information, identifying supporting evidence, and preparing messages to retailers.

Key Features

Refund case management: Create, view, and update refund cases.
Deadline tracking: Track important dates using the policy information entered by the user.
Eligibility analysis: Apply rule-based checks to help users understand their potential options.
Evidence checklist: Identify information and documents that may support a claim.
Next Best Action: Help users identify a practical next step.
Support message drafting: Generate messages that users can review before contacting a retailer.
Follow-up tracking: Record follow-up dates and case progress.
Receipt text parsing: Extract purchase details from supplied text.
Search and analytics: Find cases and review claim-management activity.
Data export: Export case information for further use.

Alexa+ Track

Refund Guardian AI includes a self-hosted Model Context Protocol (MCP) server implemented with the official Python MCP SDK. The server exposes tools for listing and creating refund cases, drafting support messages, updating case statuses, scheduling follow-ups, analyzing eligibility, and verifying case information.

The MCP server uses Streamable HTTP. Our project is designed to demonstrate how an assistant can use structured tools to help users manage refund workflows.

Technology

Python, FastAPI, SQLite, HTML, CSS, JavaScript, the Python MCP SDK, and pytest.

Current Limitations

Eligibility results depend on the information supplied by the user and are not a guarantee of a refund. Retailer policies are not independently fetched or verified. Receipt parsing works with text and is not optical character recognition (OCR). The application does not automatically submit claims or contact retailers on the user's behalf.

Our Goal

Our goal is to make refund management more organized, understandable, and actionable by turning a frustrating collection of deadlines, documents, and messages into a guided workflow.

Built With

ai
css
fastapi
html
javascript
mcp
pytest
python
sqlite

Submitted to

Build, Ship, Shape: Amazon Developer Hackathon

Created by

My Contribution to Refund Guardian AI

I contributed to the backend development of Refund Guardian AI, an AI-powered consumer-support application that helps users organize refund cases and track important deadlines.

My work focused on developing and testing REST APIs using Python, FastAPI, and SQLite. I contributed to features such as refund eligibility analysis, deadline tracking, evidence checklists, follow-up reminders, case prioritization, analytics, and search and filtering.

I also worked on CSV import and export functionality, receipt text parsing, support-message drafting, and MCP tool integration. These features help users manage refund cases more efficiently and keep their supporting information organized.

Throughout development, I focused on API validation, error handling, edge cases, and automated testing to improve reliability. Working on this project helped me strengthen my backend development, debugging, API design, and testing skills while collaborating with my teammate to build a practical solution to a real-world problem.

I’m proud to have helped turn the project idea into a functional backend that supports the core refund-management workflow.

Skanda Bhat
Prajwal Devadiga