CareLoop: The AI Caregiving Companion

CareLoop turns medication reminders into coordinated care with AI-powered follow-ups, family coordination, smart escalations, and human-approved medication management.

Comment

Inspiration

Medication reminders can notify someone, but they cannot ensure that someone follows up. When an elderly family member misses a dose, relatives may all receive the same alert, leaving everyone uncertain about who should act. We wanted to solve this gap by building an AI caregiving companion that turns reminders into coordinated action while keeping humans in control.

What it does

CareLoop is an AI-powered caregiving agent built around a simulated Alexa+ experience and a self-hosted MCP server. It tracks medication schedules, detects missed doses, alerts family members, and escalates unclaimed alerts to a backup contact after 15 minutes. When someone claims responsibility, the rest of the family is notified and escalation stops. It also logs doses, generates weekly adherence summaries, learns medication routines, proposes schedule adjustments and refills, and reads medicine bottle photos to suggest adding medications. Changes require human approval, and pharmacy orders are simulated.

How we built it

We built CareLoop using Python, the MCP Python SDK with Streamable HTTP, SQLite, Starlette, and vanilla JavaScript. Google Gemini powers the AI agent, using tool calls to interact with the caregiving system and coordinate multi-step workflows. Amazon SNS supports SMS and email notifications when configured.

Challenges we ran into

Designing reliable coordination was more complex than sending notifications. We needed to prevent duplicate alerts, stop escalation when a family member accepted responsibility, and ensure that AI-generated suggestions never became unapproved changes. Integrating MCP tools with the agent also required careful handling of tool responses and SDK version differences. Another challenge was balancing automation with safety, especially for medication-related decisions. We addressed this by enforcing critical restrictions at the tool level rather than relying solely on the AI model.

Accomplishments that we're proud of

We built a functional caregiving workflow that goes beyond reminders to coordinate responsibility and follow-up. CareLoop combines multiple MCP tools, persistent state, AI-driven decisions, family escalation, and human-approved proposals in one system. We designed the system so that the AI can recommend actions without independently approving medication changes.

What we learned

We learned how MCP enables AI agents to interact with real application capabilities through structured tools, and how an agent can coordinate multi-step workflows instead of merely answering questions. We also learned that reliable AI applications need deterministic guardrails, persistent state, clear explanations, and human oversight—particularly in sensitive domains such as caregiving.

What's next for CareLoop: The AI Caregiving Companion

We want to integrate with real Alexa+ experiences and compatible smart-home devices, improve personalized medication routine learning, and strengthen reliability through testing and monitoring. Future versions could support caregiver dashboards, configurable escalation policies, and integrations with pharmacies and healthcare providers.

Our goal is to make CareLoop a dependable caregiving companion that helps families stay informed, share responsibility, and provide better support without taking important decisions away from people.

Built With

Share this project:

Updates