Alexa+ Enterprise Hub

An autonomous supply chain assistant leveraging MCP to provide small businesses with predictive stockouts, budget planning, and automated purchase orders via conversational AI.

Comment

Inspiration

Small shopkeepers, hawker stalls and local cafe owners constantly fight a battle against dead capital and inventory spoilage. They rely on pen-and-paper tracking, making it incredibly difficult to predict stockouts or optimize reorder budgets. I wanted to bring enterprise-grade supply chain intelligence to local vendors. The idea was simple which was to make complex logistics as easy as having a conversation with an assistant.

What it does

The Alexa+ Enterprise Hub is a predictive Small Business Inventory & Order MCP Server. It acts as an autonomous supply chain assistant. Key capabilities include:

  • Predictive Stockouts: Tracks dynamic daily burn rates and warns the owner before high-velocity staples run out.

  • Smart Budget Planning: Simulates the most critical reorders needed to stay operational under strict cash-flow constraints.

  • Automated Purchase Orders: Aggregates low-stock items by supplier and generates dispatch-ready payloads.

  • Clearance Engine: Scans for near-expiry items and generates promotional discount strategies to recoup costs.

    How we built it

    To fulfill the Alexa+ Track requirements, I built a self-hosted Model Context Protocol (MCP) server that strictly adheres to the 2025-11-25 Streamable HTTP specification.

Backend: Powered by Python, FastAPI, and SQLite, exposing 15+ highly specialized enterprise tools via JSON-RPC.

Frontend Simulation: Because I am new to MCP, I utilized the allowed "simulated Alexa+ experience" path. I built an embedded, browser-based Agentic Chat Simulator that executes complex MCP tool invocations natively from a clean dashboard.

Mathematical Core: The backend utilizes algorithmic thresholds to automate logic.

Challenges we ran into

Schema Migration Friction: While expanding the MCP tool definitions to support new financial calculations, I added a unit_cost column to the Python database initialization. I expected SQLite to automatically append the new schema, but the application threw persistent 500 Internal Server Errors. I resolved this by manually clearing the local .db file to force a clean recreation, highlighting the need for built-in migration utilities in local-first AI apps.

Natural Language Parsing: Initially, the simulated chat client used basic regex to parse complex commands like "Add new item KFC Bucket, Fast Food, 15...". The parser frequently broke on missing fields. I pivoted to building a dedicated structured GUI form that bypasses the text parser and sends a perfectly formed JSON-RPC payload directly to the MCP endpoint, ensuring deterministic reliability.

Accomplishments that we're proud of

Mastering the MCP Specification: I am incredibly proud of successfully building a self-hosted Model Context Protocol (MCP) server from scratch that fully complies with the latest 2025-11-25 Streamable HTTP specification. Decoupling the complex backend supply-chain logic from the conversational UI was a huge milestone.

Building a Seamless Alexa+ Simulator: Since I am new to MCP, creating a robust, browser-based Agentic Chat Simulator was a major achievement. I successfully bridged the gap between natural language commands and strict JSON-RPC schemas, allowing users to interact with the tools natively in the browser without needing to configure external desktop clients.

Enterprise Logic for Local Shops: I successfully translated complex enterprise-grade supply chain mathematics like dynamic daily burn rates, safe runway thresholds, and real-time profit margin projections into a lightweight, lightning-fast SQLite architecture that any small shopkeeper can run locally.

Resilient User Experience: When I realized that basic text parsing was too brittle for adding complex custom items (which required 9 distinct data points like expiry dates and supplier names), I quickly pivoted. I am proud of engineering a hybrid UI that combines the magic of a conversational agent with the rock-solid reliability of direct GUI-to-MCP JSON payloads.

What we learned

Building to the MCP specification was a paradigm shift. I learned how standardizing tool-calling via JSON-RPC completely decouples business logic from the conversational UI. It makes the backend incredibly scalable and agnostic to whichever client or LLM is consuming it.

What's next for Alexa+ Enterprise Hub

I plan to integrate real-world wholesale supplier APIs so that the "Generate Purchase Order" tool actually dispatches an email or EDI payload to distributors. I also want to port the client-side simulator into a native Fire OS application to bring this dashboard directly to storefront displays.

Built With

Share this project:

Updates

Submission history