Alexa+ Concierge: Streamable MCP Bridge

Self-hosted Streamable HTTP MCP server and web simulator for agentic Alexa+ tools and AWS automation.

Comment

With Alexa+ adopting the open Model Context Protocol (MCP), developers now have the ability to expose dynamic agentic capabilities without legacy skill scaffolding. Our goal was to create a lightweight, responsive bridge that allows Alexa+ agents to discover services, run diagnostic health queries on smart home hardware, and trigger security routines in real time over Streamable HTTP.

What It Does

Alexa+ Concierge is an implementation of a self-hosted MCP server conforming to the 2025-11-25 Streamable HTTP specification. It features:

  • MCP Discovery & Transport: Exposes standard endpoints (/mcp/tools and /mcp/call) enabling agentic runtimes to discover and invoke tools.
  • Device Telemetry Checks: Queries device state, battery health, and round-trip network metrics.
  • Automated Security Protocols: Executes multi-zone access lockdowns designed along AWS AgentCore design patterns.
  • Interactive Simulator: A built-in web frontend allowing immediate inspection and verification of agent invocations and JSON payloads.

How We Built It

  • Language & Runtime: Python 3.11 with aiohttp to manage asynchronous HTTP streaming and REST routing.
  • Protocol: Model Context Protocol (MCP) streamable transport specification.
  • Architecture: AWS AgentCore and Bedrock dispatch-compatible schemas for tool binding and zero-latency execution.
  • UI & Inspection: Lightweight vanilla HTML5/CSS and JavaScript console logging tool calls and responses in real time.

Challenges We Ran Into

Adhering to the November 2025 Streamable HTTP transport format while maintaining an ultra-lightweight footprint without external complex agent runtimes. Ensuring local asynchronous event loops handled rapid tool dispatches cleanly was also a key focus.

Accomplishments We're Proud Of

  • Complete zero-dependency implementation of the MCP Streamable HTTP specification.
  • A functional web-based simulation environment that requires only a single Python command to run and demonstrate.
  • Seamless compatibility with both Alexa+ agentic patterns and AWS Builder tool standards.

What We Learned

We gained a deep practical understanding of the Model Context Protocol, the shift from traditional intent-based voice models to decoupled agentic tool calling, and how streamable endpoints facilitate multi-turn agent interactions.

What's Next

  • Direct streaming integration into AWS Bedrock agents via AWS Lambda.
  • Expanding tool endpoints to interface with physical Ring and Bee IoT telemetry endpoints.## Inspiration

Built With

Share this project:

Updates

Submission history