Aura+

Inspiration

The advent of Alexa+ marks the single biggest transition in smart living: moving away from rigid voice intents and hard-coded syntax toward fluid, conversational intelligence powered by Foundation Models and open agentic protocols. However, today's smart homes remain deeply fragmented—cameras exist in one silo, thermostats and lighting in another, and grid energy telemetry in a third.

We asked ourselves: What if Alexa+ could act as a true ambient consciousness for the home? An autonomous orchestrator capable of reasoning across multi-modal optical feeds (Ring), grid tariffs, and ambient living preferences—powered by the open Model Context Protocol (MCP) and Amazon Bedrock.

What it does

Aura+ is an ambient intelligent operating system and self-hosted MCP server designed specifically for Alexa+. It features:

Self-Hosted Streamable HTTP MCP Server (Spec 2025-11-25+): Exposes 5 production tools across home actuation, Ring security triage, calendar coordination, and power grid optimization via Server-Sent Events (SSE) and JSON-RPC 2.0.
Multi-Modal Ring Sentry Triage: Connects to Ring optical feeds (Doorbell Pro 2, Floodlight Cam) to identify delivery drivers, analyze threat postures, and automate intelligent responses without false alarms.
Eco-Pulse Energy Peak Shaving: Real-time power load balancing that dynamically sheds 25–35% of non-essential wattage (lighting zones, HVAC buffers) ahead of utility peak pricing windows.
Echo Show / Ambient Portal Experience: A futuristic, glassmorphism web console featuring a living Alexa+ voice orb, bidirectional speech recognition & synthesis, and an interactive MCP Inspector Drawer that lets judges observe live SSE streaming tool calls in real time.

How we built it

Protocol Layer: Built an Express MCP Server complying strictly with the 2025-11-25+ Streamable HTTP specification, supporting bidirectional JSON-RPC 2.0 and SSE lifecycle sessions (/mcp/sse, /mcp/messages, /mcp/rpc).
AI & Agent Core: Integrated Amazon Bedrock via @aws-sdk/client-bedrock-runtime using amazon.nova-pro-v1:0 and anthropic.claude-3-5-sonnet for multi-step agentic planning. Includes an intelligent zero-config local simulation fallback so anyone can evaluate the project out-of-the-box.
Ambient Frontend: Developed with React and modern Vanilla CSS utilizing glassmorphism, responsive grid layouts, dynamic SVG rotating orbs, and Web Speech API vocalization.
Device Telemetry: Implemented real-time state synchronization for multi-zone lighting, Ultion smart deadbolts, Ecobee/Alexa thermostats, and solar inverter telemetry.

Challenges we ran into

Handling real-time streaming JSON-RPC over Streamable HTTP SSE connections while ensuring seamless backward compatibility with stateless RPC callers.
Designing multi-agent tool execution flows where a single natural language request (e.g. "Alexa, goodnight") cleanly dispatches compound tool calls (locking doors, arming Ring sentries, adjusting climate buffer) without hallucinated tool arguments.

Accomplishments that we're proud of

Complete implementation of the November 2025 Streamable HTTP MCP standard with zero external middleware dependencies.
A breathtaking, ultra-responsive ambient display that makes complex agentic AI and protocol inspection immediately intuitive to anyone.
Seamless zero-config execution: the project can be cloned and run with npm start in under 60 seconds.

What we learned

The Model Context Protocol (MCP) is remarkably versatile for consumer IoT and ambient AI—it transforms Alexa+ into an extensible agentic platform that any developer or brand can build for.
Combining Amazon Bedrock Foundation Models with MCP tool calling eliminates the need for complex, brittle rule-based automation engines.

What's next for Aura+

Deploying Aura+ as a native AWS Lambda container accessible via public Amazon API Gateway.
Direct integration with Amazon Sidewalk and Matter/Thread mesh networks.
Expanding multi-modal computer vision directly on edge devices using AWS IoT Greengrass.

Instructions for Testing & Evaluation

Clone the repository: git clone https://github.com/dhanrajgupta2736/amzndevchalen.git
Run automated test suite: npm test (passes all 6 specification checks).
Start the application: npm start (opens on http://localhost:3001).
Click any of the 4 1-Click Demo Scenarios along the top bar or open the ⚡ MCP Inspector drawer to inspect live tool calls.