Connect Claude and ChatGPT to Health OS

Health OS exposes an "agent seam" — an MCP server that lets Claude or ChatGPT read your metrics and run N-of-1 experiments right inside the conversation. Connecting takes a couple of minutes and stays fully under your control: revoke the token any time, and review or undo every write in Settings.

What it does

  • Reads your metrics: sleep, HRV, heart rate, activity, screen time — from Oura, RescueTime and manual entries.
  • Starts and runs protocols: kicks off an N-of-1 experiment, logs phases, tells you today's step.
  • Logs data from what you say: mood, energy, workouts — without opening the app.

Everything runs on Google Cloud infrastructure in the EU (europe-west1).

Connect Claude

  1. Open claude.ai → Settings → Connectors.
  2. Click "Add custom connector".
  3. Enter the name "Health OS" and the URL https://os.health-os.app/mcp.
  4. Click "Connect", sign in with Google, and allow access.

Custom connectors aren't available on every Claude plan — check that yours supports them.

Connect ChatGPT

  1. Open ChatGPT → Settings → Connectors and turn on developer mode.
  2. Add a connector by URL: https://os.health-os.app/mcp.
  3. Sign in with Google and allow access.

Availability depends on your plan and region — ChatGPT connector apps may not yet be available in the EEA.

Try these prompts

Show my HRV trend for the last 30 days and tell me if it is outside my usual range.

Start the Health OS protocol "Caffeine after 2 PM".

What is today's step of my active protocol?

Log energy 4 and mood 3.

Compose a 20-minute bodyweight workout and log it when I'm done.

Summarise my last experiment honestly — is the effect bigger than noise?

Privacy & control

  • The assistant sees only what the MCP tools return — not your whole account.
  • Revoke the token any time in Settings → API tokens.
  • Every write from the assistant is stamped via: agent and can be undone in Settings → Recent changes.
  • Health OS stores and processes your data in the EU. Whatever the tools return into the conversation is received by the assistant you chose (Anthropic or OpenAI) and handled under its own terms.
Privacy Policy →

Limits

  • Health OS is not a medical device and does not replace a physician.
  • N-of-1 results are personal and noisy: one experiment on one person is an observation, not proof.
  • The assistant relies on your data and phrasing — double-check anything that matters.