BLUEPRINTAI Center of Excellence

Vital Intelligence Health System

Wellness-logging assistant that summarizes wearable signals for personal planning context

Advanced6 weeks~$400/mo

Multi-Cloud

The Problem

AI assistants ignore the human operating context. They schedule deep work after short sleep, push notifications during planned downtime, and treat every day as identical. Wearable data can support better self-reflection when it is handled conservatively and with clear health boundaries.

The Solution

Deploy a Vital Intelligence MCP server that ingests wearable data via HealthKit/Oura/Whoop APIs, computes a personal context summary, and exposes tools that any Claude agent can call: get_energy_context, suggest_task_timing, summarize_stress_notes, log_recovery_notes. The agent becomes wellness-context aware without offering medical advice.

Overview

A wellness integration system that connects wearable devices (Apple Watch, Oura Ring, Whoop) to an AI assistant for personal planning context. The system ingests user-authorized signals — sleep duration, heart rate variability, activity, and recovery notes — and summarizes patterns for reflection. It is a logging and education layer, not a medical device, diagnosis tool, or treatment recommendation system.

Architecture

Loading interactive diagram...

Components

Wearable Data Ingest

external

Connects to Apple HealthKit, Oura API, Whoop API, and Garmin Connect. Pulls sleep, HRV, resting HR, activity, and recovery data.

Service: OAuth2 integrations

Vitals Processor

compute

Normalizes biometric data from multiple sources. Computes rolling averages, trend detection, and anomaly flags.

Service: Railway (Node.js)

Wellness Context Model

ai-service

Summarizes wearable and self-reported signals into planning context: recent sleep, subjective energy, activity trend, and user-noted recovery needs. Uses Claude for plain-language summaries, not diagnosis.

Service: Claude Haiku + custom model

Vitals Time-Series DB

database

Stores biometric history with per-user isolation. Time-series optimized for trend queries and pattern detection.

Service: Supabase (TimescaleDB extension)

Vital Intelligence MCP

gateway

MCP server exposing wellness-context tools to any Claude agent. Tools: get_energy_context, suggest_task_timing, summarize_stress_notes, log_recovery_notes, get_sleep_summary.

Service: Railway (MCP SDK)

Adaptive Scheduler

compute

Reorders task queue based on cognitive state. Moves deep work to high-energy windows, admin to low-energy periods.

Service: Vercel Cron + Calendar API

Wellness Notes

ai-service

Generates educational prompts and self-reflection notes: breathing practice reminders, movement-break ideas, hydration check-ins, and user-authored recovery logs.

Service: Claude Sonnet

Health Dashboard

gateway

Real-time vitals dashboard showing energy levels, sleep trends, stress patterns, and AI adaptation history.

Service: Next.js on Vercel

Implementation Steps

Wearable Integration

2 weeks

Connect to health data sources and build the ingest pipeline

Tasks

Implement Apple HealthKit OAuth flow
Add Oura Ring API connector
Add Whoop API connector
Build data normalization layer (unified schema)
Deploy time-series storage on Supabase

Deliverables

Multi-wearable data pipelineUnified biometric schema

Cognitive State Model

2 weeks

Build the model that maps biometrics to cognitive states

Tasks

Define cognitive state dimensions (energy, focus, creativity, stress)
Build scoring algorithm from biometric inputs
Train pattern detection on personal history
Implement anomaly detection (unusual HRV, poor sleep)
Create adaptive thresholds per user

Deliverables

Cognitive state scoring enginePersonal baseline calibration

MCP Server & Interventions

2 weeks

Deploy the MCP server and wellness intervention system

Tasks

Build Vital Intelligence MCP server (5 tools)
Deploy on Railway with auto-scaling
Implement adaptive scheduling integration
Create wellness intervention library (breathing, movement, hydration)
Build health dashboard with real-time charts

Deliverables

Production MCP server on RailwayHealth dashboardIntervention system

Code Examples

Vital Intelligence MCP Server — Health Tools

MCP server that exposes biometric-aware tools to any Claude agent

import { Server } from '@modelcontextprotocol/sdk/server'
import { StdioTransport } from '@modelcontextprotocol/sdk/server/stdio'

const server = new Server({ name: 'vital-intelligence', version: '1.0.0' })

server.tool('get_energy_level', 'Current energy level based on sleep, HRV, and activity', {
  type: 'object', properties: {}, required: []
}, async () => {
  const vitals = await getLatestVitals(userId)
  const energy = computeEnergyScore(vitals)
  return {
    content: [{
      type: 'text',
      text: JSON.stringify({
        score: energy.score, // 0-100
        level: energy.level, // 'high' | 'medium' | 'low' | 'critical'
        factors: {
          sleep: vitals.sleepScore,
          hrv: vitals.hrvTrend,
          activity: vitals.activityLevel,
        },
        recommendation: energy.recommendation,
      })
    }]
  }
})

server.tool('suggest_task_timing', 'Optimal time windows for different task types', {
  type: 'object',
  properties: { taskType: { type: 'string', enum: ['deep-work', 'creative', 'admin', 'exercise'] } },
  required: ['taskType']
}, async ({ taskType }) => {
  const schedule = await getOptimalWindows(userId, taskType)
  return { content: [{ type: 'text', text: JSON.stringify(schedule) }] }
})

const transport = new StdioTransport()
await server.connect(transport)

Cost Estimate

$400

per month

$4,800

per year

Railway (MCP server)

$100

Supabase Pro

$100

Claude API (scoring)

$100

Vercel (dashboard)

$100

Assumptions: 1 user (personal), ~100 biometric syncs/day, 5 MCP tool calls/hour

Use Cases

Personal health-aware AI assistantExecutive performance optimizationAthlete recovery and training AIWellness coaching platformBurnout prevention system

Technologies

MCP SDKApple HealthKitOura APIWhoop APIRailwaySupabaseTimescaleDBClaude APINext.jsChart.jsTypeScript

Ready to Build?

Deploy this architecture in minutes, or get the production-ready template with full source code.

Try Prototype Get Template