Kiso (基礎)

The backend platform for Kasane and Milo.

Kiso is a health intelligence server that powers two clients: the Kasane iOS app (habits, coaching, focus plans) and Milo (an AI coaching agent on WhatsApp). Both read and write the same person profiles, habits, check-ins, and health data through a shared API.

The name means "foundation" in Japanese. Kasane means "layers." Kiso is what the layers sit on.

What it does

For the iOS app: bidirectional sync of persons, habits, check-ins, focus plans, and health measurements via REST API.

For Milo: 52 MCP tools for health coaching. Log weight, meals, labs, blood pressure. Pull data from Garmin, Apple Health, Oura, Whoop. Score 20 health metrics against NHANES population percentiles and clinical guidelines.

For both: a unified person context that merges all data sources in one call.

Quick start

git clone https://github.com/a-deal/kiso.git
cd kiso
pip install -e ".[gateway,dev]"

# Run the gateway
python3 -m uvicorn engine.gateway.server:create_app --factory --port 18800

# Run tests
python3 -m pytest tests/ -v

MCP server (for Claude Desktop / Claude Code)

{
  "mcpServers": {
    "kiso": {
      "command": "uvx",
      "args": ["kiso"]
    }
  }
}

API surfaces

Full API reference: docs/API.md

Architecture

One Python process. Two storage systems. Three clients.

• SQLite (data/kasane.db): persons, habits, check-ins, focus plans, wearable daily metrics, labs, meals, weight, training. Synced with iOS and wearable integrations.
• CSVs (data/): legacy weight, meals, labs, supplements. Being migrated to SQLite.
• Bridge: person.health_engine_user_id links a SQLite person to a per-user data directory.

Full architecture: docs/ARCHITECTURE.md

Health scoring

20 clinically validated metrics scored against NHANES population data (300K+ Americans) and clinical guidelines from AHA, ADA, and ESC. Covers cardiovascular, metabolic, body composition, recovery, and lifestyle dimensions.

Going from 0% to full coverage costs under $300.

Full methodology: docs/METHODOLOGY.md

Docs

License

MIT