Nexa

Nexa gives Alexa+ household context by connecting people, visitors, deliveries, and tasks into one safe, conversational experience powered by MCP and Amazon Bedrock.

Comment

Inspiration

Homes are full of useful information, but that information is usually scattered across messages, reminders, delivery notifications, calendars, and different people.

We wanted to explore what a truly context-aware household assistant could feel like — not just an assistant that responds to commands, but one that understands what is happening across the home.

That idea became Nexa: a household intelligence experience designed for Alexa+.

What it does

Nexa brings together household context around people, visitors, deliveries, and tasks into one conversational experience.

Users can ask questions like:

  • “Who is home?”
  • “Who am I expecting today?”
  • “What deliveries are pending?”
  • “Give me a home update.”

Nexa can also take actions, such as updating household presence or managing tasks. Actions that change household state require explicit confirmation before they are executed.

The result is an experience that moves from:

State → Context → Safe Action

while keeping household information persistent.

How we built it

Nexa is built as a production FastAPI application deployed on Amazon ECS.

We created a self-hosted MCP server using protocol version 2025-11-25 with Streamable HTTP, exposing 14 household tools for capabilities including:

  • Household presence
  • Visitors
  • Deliveries
  • Tasks
  • Household briefings

For contextual household summaries, Nexa uses Amazon Bedrock with Amazon Nova Lite.

Household state is stored persistently in Amazon RDS for PostgreSQL, while AWS Secrets Manager, Amazon CloudWatch, and AWS IAM support configuration, monitoring, and access control.

Because direct Alexa+ developer integration was not available for hackathon participants, we built a custom simulated Alexa+ experience to demonstrate how Nexa could work as an Alexa+ household experience.

Challenges we ran into

One of the biggest challenges was turning several independent household data sources into a single coherent conversational experience without making the system feel like a collection of disconnected APIs.

We also had to design safe write operations. Reading household information is straightforward, but changing state — such as updating someone’s presence or creating a task — needed a predictable confirmation flow before anything was persisted.

On the infrastructure side, we worked through production deployment issues involving:

  • ECS configuration
  • Database connectivity
  • AWS Secrets Manager
  • IAM permissions
  • Amazon Bedrock access
  • Persistent state validation

Testing MCP over Streamable HTTP also required making sure the protocol implementation worked reliably both locally and against the deployed service.

Accomplishments that we're proud of

We are especially proud that Nexa is not just a visual prototype.

The project includes:

  • A working production deployment on AWS
  • A self-hosted MCP server with 14 household tools
  • Amazon Bedrock integration using Nova Lite
  • Persistent PostgreSQL household state
  • Confirmation-gated state-changing actions
  • A simulated Alexa+ experience
  • A tested conversational workflow connecting household state, context, and action

The most important accomplishment is that the demo reflects real system behavior: when Nexa changes household state, that change persists and affects subsequent responses.

What we learned

We learned that contextual assistants become much more useful when they can combine multiple small pieces of state instead of treating every request as an isolated interaction.

We also learned that agentic functionality needs clear boundaries. The ability to act is valuable, but predictable confirmation and deterministic execution are essential when an assistant is changing real household information.

Building with MCP reinforced the value of exposing capabilities as structured tools rather than tightly coupling every action to the conversational layer.

Finally, deploying the full experience on AWS showed us how important infrastructure, permissions, persistence, monitoring, and failure handling are when moving an AI prototype toward a real product.

What's next for Nexa

The next step is to make Nexa more proactive and more useful over time.

We want to expand the experience with features such as:

  • “What changed while I was away?”
  • “Does anything need my attention?”
  • Household activity timelines
  • Morning, leaving-home, and evening briefings
  • Smarter suggestions based on household context
  • Better explanations of why Nexa surfaced a particular update
  • More household integrations and devices
  • Richer multimodal experiences for Alexa+ screens

The long-term vision is for Nexa to become a trusted household intelligence layer that understands what is happening across the home and helps people act on that context naturally and safely.

Built With

Share this project:

Updates

Submission history