OmniFlow: Agentic Alexa+ Life & Commerce Copilot
OmniFlow is an autonomous, multi-modal Alexa+ Agent Skill and self-hosted Model Context Protocol (MCP) Server compliant with the latest MCP specification version 2025-11-25 over Streamable HTTP, integrated with Amazon Bedrock.
Instead of simple one-shot Q&A, OmniFlow orchestrates complex multi-step daily workflows across services: researching travel/events, analyzing budgets, finding matching Amazon products, and generating interactive visual cards (media carousels, action triggers, and one-click approvals) for screen-enabled Alexa+ and web clients.
💡 Inspiration
Most voice assistants remain reactive "smart speakers" stuck in single-turn command loops: "Alexa, what is the weather?" or "Alexa, add milk." With the advent of Alexa+ and Model Context Protocol (MCP), assistants can graduate from simple query-response bots to autonomous goal executors. When a user says:
"I'm planning a 3-day weekend camping trip to Olympic National Park under $300 next Friday, check weather gear, book camping permits, and order needed supplies."
OmniFlow breaks this high-level intent down into actionable agentic sub-goals, calls specialized tools over Streamable HTTP MCP, synthesizes optimal decisions, and returns structured rich-media cards for instant user confirmation.
🚀 Key Features
- Alexa+ Agent Skill & MCP Server: Implements the Model Context Protocol (MCP specification version 2025-11-25) over Streamable HTTP transports.
- Multi-Modal Visual UI: Renders dynamic cards, product carousels with Prime badge metadata, itinerary timelines, and interactive confirmation buttons in an Alexa+ simulator web app.
- Amazon Bedrock AI Engine: Leverages Claude on Amazon Bedrock for structured reasoning, function calling, and multi-turn conversational memory.
- Autonomous Multi-Step Orchestration:
- Travel & Event Scheduler Tool: Queries destinations, check-in schedules, and weather patterns.
- Amazon Commerce & Cart Optimizer Tool: Searches Amazon products, compares reviews/pricing, and builds optimized shopping baskets.
- Smart Budget & Action Guardrail Tool: Enforces cost ceilings and requires human-in-the-loop explicit confirmation before execution.
- Context-Aware Cross-Session Memory: Remembers user preferences, dietary limits, and past purchases across sessions.
🛠️ How We Built It
- Frontend / Simulator: Built with modern responsive UI with voice wave animations, speech synthesis & recognition, and interactive cards.
- Backend / MCP Server: Fast Node.js Streamable HTTP server exposing standard MCP endpoints (
/mcp,/mcp/stream,/mcp/info) conforming to the official 2025-11-25 MCP spec. - AI Agentic Layer: Amazon Bedrock SDK integration running an agentic tool-use loop with streaming responses and JSON-schema validated parameters.
🧗 Challenges We Faced
- Adapting the newly standardized MCP Streamable HTTP protocol to low-latency streaming user experiences.
- Ensuring seamless fallback when Bedrock tools run in parallel vs sequential execution pipelines.
- Designing an intuitive multi-modal UI where audio responses match visual card states without visual clutter.
🏆 Accomplishments that we're proud of
- Successfully implemented full MCP spec 2025-11-25 streamable HTTP protocol.
- True multi-step autonomous execution: zero manual glue required between travel planning and Amazon product checkout preparation.
- Production-ready developer experience with tests, clear documentation, and a detailed friction log.
Built With
- agentic-ai
- alexa
- amazon-bedrock
- amazon-web-services
- mcp
- model-context-protocol
- node.js
- react
- tailwind-css
- typescript

Log in or sign up for Devpost to join the conversation.