Inspiration

Inspiration

With the evolution of AI agents, standardizing agentic tool access is critical. The Model Context Protocol (MCP) spec (2025-11-25) provides a robust specification for secure tool discovery and execution over Streamable HTTP. We wanted to build a seamless bridge between modern LLMs (AWS Bedrock Claude 3 Haiku) and custom local/cloud agent capabilities, complete with a glassmorphic simulator for real-time interaction.

What it does

  • Alexa+ Track Compliance: Implements full MCP specification featuring Streamable HTTP transport and dynamic tool execution (get_weather, control_smart_device).
  • AWS Bedrock Integration: Uses AWS Bedrock Runtime SDK with Claude 3 Haiku for ultra-fast, intelligent reasoning and structured JSON tool calls.
  • Interactive Simulator: A React + Vite + Tailwind CSS frontend that simulates device voice/text interactions, displaying active tool execution status, payload inspects, and real-time streaming state.

How we built it

  • Backend: Node.js, Express.js implementing streamable HTTP endpoints for MCP discovery (/mcp/tools) and execution (/mcp/execute).
  • Cloud AI: Integrated @aws-sdk/client-bedrock-runtime using Bedrock Converse API for structured tool use.
  • Frontend: React, Tailwind CSS, Lucide React icons, Vite.
  • Documentation: Included comprehensive FRICTION_LOG.md for ecosystem improvement feedback.

Challenges we ran into

  • Ensuring strict compliance with the newly updated 2025-11-25 MCP spec while handling asynchronous Streamable HTTP tool calls without blocking state updates.
  • Structuring strict tool definition JSON schemas for AWS Bedrock Converse API.

Accomplishments that we're proud of

  • Successfully executing zero-latency end-to-end tool calls through AWS Bedrock.
  • Building a sleek, highly responsive glassmorphic UI simulator.

What we learned

  • Deep dive into Model Context Protocol internals.
  • Practical orchestration of agentic tools with AWS Bedrock SDK. ## What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for Alexa+ MCP Agent & Web Simulator

Built With

Share this project:

Updates

Submission history