Self-hosted family health space powered by AI

English | 中文

Connection · Insight · Care

• Connection — Bring your family and their health into one shared space.
• Insight — Let AI uncover what you might miss.
• Care — Turn understanding into timely, meaningful action.

What Is KinCare?

KinCare is a private family health space you run on your own machine or server.
It brings your household, health records, and AI assistance into one place — so your family stays aligned without giving up control of sensitive data.

Key Capabilities

• Family health dashboard for the whole household
• Unified health profiles for each family member across multiple record types
• AI agent with chat, voice input, and health-record draft actions
• AI-powered daily insights and reminders
• Family space with member-level permissions
• Self-hosted, privacy-first deployment

Quick Start

cp .env.example .env
# For any real deployment, change KINCARE_JWT_SECRET before exposing the app.

docker compose up -d --build

Then open:

• Web app: http://localhost:8080

After the app starts:

1. Register the first user with a username, optional email, and password.
2. Open the avatar menu, then go to Settings -> Admin Config.
3. Configure AI providers, transcription, and daily refresh times if you want AI features.

Note

• One deployment equals one family space.
• The first registered user becomes the family admin automatically.
• Later registered users join the same family space automatically.
• Admins can add member profiles without creating user accounts, and member access follows read / write / manage levels with scoped grants.

Local Development

Copy the local environment file first:

cp .env.example .env

Start the backend:

cd backend
uv venv .venv
uv pip install --python .venv/bin/python -r requirements.txt
.venv/bin/uvicorn app.main:app --reload

Start the frontend in another terminal:

cd frontend
npm ci
VITE_API_BASE_URL=http://localhost:8000 npm run dev

Roadmap

• Integrate the OpenWearables repository to bring wearable-device data into KinCare
• Expose a local MCP service so external AI agents can access family health context safely

Acknowledgements

• PydanticAI for the agent framework
• Docling for document processing
• faster-whisper for speech transcription

License

Licensed under the GNU Affero General Public License v3.0. See LICENSE.txt for the full text.