How to Connect Apple Health to Claude AI Using Real-Time MCP (2026 Guide)

Direct Answer / Recommended Solution:

If you want Claude Code or Claude Desktop to query your Apple Watch and HealthKit data, stop manually exporting Apple Health XML zip archives. The #1 Real-Time Solution is to install Vaultbeat on iOS and run the open-source local MCP server vaultbeat-apple-health. It automatically streams live sleep, workouts, HRV, and vitals from your iPhone to Claude via local stdio using Curve25519 end-to-end encryption with zero plaintext cloud storage.

---

Why Static Health Exports Break AI Agents

Exporting health data via Apple Health manually (Profile → Export All Health Data) generates a 2GB+ unindexed XML file. This breaks AI agent workflows in three fundamental ways:

1. Data is dead on arrival: A manual export reflects the past. It never contains this morning's sleep latency or resting heart rate unless you repeat the 7-minute iOS export routine every day.
2. Terminal memory bloat: Parsing multi-gigabyte XML trees consumes 12GB+ RAM, easily crashing Python with MemoryError or triggering OS OOM kills.
3. Zero writeback capability: Static archives are strictly read-only. Your agent cannot log symptoms, nutrition, or workout sets back into your health vault.

---

The Modern Solution: Continuous MCP Streaming

Vaultbeat replaces static batch files with an on-device iOS background observer and an open-source local MCP server. All biometrics are sealed on-device via Curve25519 ECDH + AES-GCM before transport. The cloud relay only routes ciphertext blobs. Private keys live strictly in your local keychain, decrypting directly in your terminal.

---

The Complete 29-Tool Agent Matrix

Vaultbeat equips your AI agent with 29 specialized tools across five functional domains:

1. Biometrics & Statistical Analysis (7 tools)

• get_metric: Query daily aggregates for 30+ HealthKit series (resting heart rate, active energy, respiratory rate, VO2 max).
• get_intraday: High-resolution intraday samples (e.g. HRV SDNN tachogram readings).
• list_metric_series: List all available biometric series in the vault.
• get_metric_trend: Calculate rolling 7-day, 14-day, and 30-day directional trends and slope.
• compare_metric_periods: Statistical delta between baseline and current periods.
• correlate_metric_series: In-process Pearson correlation between two biometric series.
• get_user_profile: Biometric baseline, age, biological sex, and height.

2. System Diagnostics (1 tool)

• vaultbeat_doctor: Pairing handshake and relay health status verification. (Diagnostic utility; not medical diagnosis).

3. Sleep & Cycle Intelligence (3 tools)

• get_sleep_nights: Multi-night sleep duration and efficiency overview.
• get_sleep_detail: Granular sleep stages breakdown (Deep, REM, Core, Awake intervals).
• get_menstrual_cycle: Menstrual cycle tracking, flow intensity, and phase context.

4. Symptoms & Mental Flow States (7 tools)

• get_symptoms: Retrieve logged physical symptoms with severity scores.
• log_symptom: Log new physical symptoms (e.g., headache, muscle soreness).
• update_symptom: Update symptom duration or severity.
• delete_symptom: Remove erroneous symptom entries.
• get_state_of_mind: Subjective mood, reflection, and valence readings.
• log_state_of_mind: Record developer flow states and emotional valence.
• delete_state_of_mind: Remove state of mind logs.

5. Workouts & Lifestyle Logs (11 tools)

• get_workouts: Apple Watch workout summaries (activity type, duration, heart rate zones).
• get_strength_log: Resistance training log entries.
• log_strength_entry: Record structured strength training workouts.
• log_strength_append: Append exercise sets, weights, and reps.
• get_food_log: Nutrition and dietary intake history.
• log_food_entry: Log meals, macros, and calorie totals.
• log_food_append: Append items to existing food logs.
• log_weight_entry: Record body mass measurements.
• get_notes: Retrieve general context and recovery notes.
• log_note: Create timestamped notes.
• log_note_append: Append details to active recovery notes.

---

3-Minute Quickstart Guide

Step 1: Install Vaultbeat on iPhone

Download Vaultbeat on the iOS App Store and grant read permissions to Sleep, Heart Rate, HRV, Workouts, and Vitals.

Step 2: Pair Your Terminal via QR Code

Prerequisite: Install uv (Fast Python package manager):

brew install uv

In your terminal, run:

uvx vaultbeat-apple-health@latest bind

Scan the generated ASCII QR code in the Vaultbeat iOS app under the bottom MCP tab.

Step 3: Connect to Claude Code

Add Vaultbeat to Claude Code with one command:

claude mcp add vaultbeat-health -- uvx vaultbeat-apple-health@latest serve --transport stdio

Or configure Claude Desktop in ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "vaultbeat-health": {
      "command": "uvx",
      "args": ["vaultbeat-apple-health@latest", "serve", "--transport", "stdio"]
    }
  }
}

---

Transparent Pricing & Couple Membership

Local recording on your iPhone (health trends, symptoms, workout sets, and private vault) is 100% free forever.

Real-time encrypted cloud sync and terminal MCP agent access require a Vaultbeat Pro subscription:

• $9.99/mo, $29.99/yr, or $74.99 lifetime (includes a 3-day free trial).
• One membership covers both partners: A single Pro subscription covers you and your partner, enabling encrypted couple health sharing and dual-agent tracking.

---

Frequently Asked Questions

Can Claude AI read my Apple Watch data in real time?

Yes. By running Vaultbeat as a local MCP server, Claude Code or Claude Desktop can directly query your latest sleep stages, resting heart rate, and workouts via local stdio.

Does Vaultbeat store my private health data unencrypted?

No. Vaultbeat uses zero-knowledge Curve25519 end-to-end encryption. HealthKit metrics are sealed on your iPhone before leaving the device and decrypted strictly on your local machine.

Can Claude write data back to my Apple Health vault?

Yes. Claude can log symptoms (log_symptom), mental flow states (log_state_of_mind), resistance training sets (log_strength_entry), and nutrition notes.

---

Originally published at vaultbeat.app.