Inspiration
Elderly and dependent individuals often struggle to keep up with daily needs — doctor appointments, medication schedules, emergency situations, and staying connected with family. We wanted to build a voice-first, agentic assistant that makes these tasks effortless through natural conversation.
What it does
CareVoice AI is an agentic voice assistant, built as an MCP (Model Context Protocol) server, that helps users:
- Book and manage doctor appointments
- Get medication reminders
- Trigger emergency contact alerts
- Check weather updates
- Send automatic notifications to family members
Users interact through [Alexa+ Agent Skill / a simulated voice-web interface], and the assistant orchestrates the right tool calls behind the scenes via our MCP server.
How we built it
- MCP Server: Custom-built server exposing tools for appointments, reminders, emergency contacts, weather, and family notifications
- Voice/Agent Layer: [Alexa+ Agent Skill / simulated voice-web simulator]
- AWS Integration: [AWS Bedrock for LLM inference / AgentCore for orchestration / SageMaker ]
- Backend: [Node.js / Python] API layer connecting the MCP server to AWS services
- Frontend: React-based web simulator for demo purposes
Challenges we ran into
[e.g., designing reliable MCP tool schemas, handling voice/agent latency, simulating Alexa+ behavior without partner-only dev tools, integrating AWS services securely]
Accomplishments that we're proud of
[e.g., a working end-to-end voice-to-action pipeline, seamless AWS integration, a demo that runs reliably within the first 3 minutes]
What we learned
[e.g., MCP protocol design, agentic orchestration patterns, building accessible voice UX for elderly users]
What's next for CareVoice AI
[e.g., real Alexa+ device testing once partner access opens, expanding to more caregiver-facing tools, multi-language support]
Log in or sign up for Devpost to join the conversation.