Apple Health MCP compared — own Postgres vs Health Auto Export & export-file MCP

Apple has no server-side HealthKit API. Every sync must originate from your iPhone — so the bridge you pick decides whether Claude, Cursor, or ChatGPT can reach your health data at all.

health4.ai's wedge: Apple Health → a Supabase/Postgres database you own → a local MCP server for Claude, Claude Code, Cursor, ChatGPT, and Ollama. Continuous HKObserverQuery sync — not same-WiFi phone TCP, and not a one-shot XML export.

This page compares the realistic options today: health4.ai, Health Auto Export, ChatGPT Health, neiltron/apple-health-mcp-style export+npx MCP, the claude.ai native connector, VitalTrends, and manual XML export. Ready to try the own-DB path? Start with the~15-minute Apple Health MCP setup.

Continuous BYO-Supabase vs same-WiFi HAE TCP

Health Auto Export's MCP server opens a local TCP port on your iPhone and expects Claude Code (or another MCP client) to connect directly across your LAN. That works fine when your Mac and your phone are on the same home network — but Claude Code runs on your laptop, your laptop travels, and your phone doesn't always follow. The moment you open a terminal at a coffee shop or in the office, the MCP server is unreachable and your agent loses all health context.

health4.ai syncs continuously via HKObserverQuery into a Supabase project you control. The MCP server connects to that database from anywhere with an internet connection — bring-your-own Postgres, not a phone-side TCP socket.

Export-file + npx MCP (neiltron/apple-health-mcp)

Projects like neiltron/apple-health-mcp take a one-shot Apple Health XML/CSV export and wrap it in an MCP server you can often start with npx. That is a fine weekend project: zero iOS app, fast to try, and useful for a static snapshot.

The trade-off is freshness. You re-export whenever you want new data; there is no continuous HealthKit observer and no always-on Postgres of your own. health4.ai is the opposite bet — an iOS app that keeps syncing into your Supabase so Claude/Cursor agents always see recent samples. We do not publish an npx one-liner yet; install is still clone + Python (see Apple Health MCP setup).

ChatGPT Health

OpenAI's ChatGPT Health feature is a cloud product for ChatGPT (US availability, product details evolve). It is not Codex, not a local MCP server, and not something you point Claude Code or Cursor at. If your goal is chatting inside ChatGPT about health, it may be enough.

If your goal is agent tooling — Claude Code, Cursor, custom MCP clients — you still need a connector that speaks MCP against data you control. That is the niche health4.ai targets.

Where health4.ai isn't the right choice

We don't have Android support yet — HealthKit is Apple-only and we haven't built a Google Fit or Health Connect bridge. If your health data lives on an Android device, none of the iOS-native options here apply; manual CSV export from your wearable's app is currently your most practical path to an AI agent.

health4.ai is a connector, not a dashboard — you use it through Claude, ChatGPT, or Cursor rather than a built-in app screen. If you want charts, trends, and a polished visual UI, VitalTrends or the claude.ai native connector will feel more complete out of the box.

The iOS app is invite-only TestFlight today — not a public App Store launch. Join the waitlist and request a beta invite; there is no public TestFlight URL.