OpenVitals — your Google Health data, as an MCP server for your AI

The problem

Asking an AI about your own health data shouldn't cost a subscription

Google will happily analyse your health data for you — inside a paid product, on their servers. As one person, you shouldn't need a monthly plan just to ask "how did I sleep this week versus last?" OpenVitals hands that job to the AI you already use.

	A paid health-AI product	OpenVitals
Cost	Monthly subscription	Free — uses the AI you already have
Your data	Leaves your device	Stays in a SQLite file on your box
Credentials	Their enterprise setup	Your own free Google Cloud OAuth client
The AI	Locked to their product	Any MCP client — Claude, ChatGPT, Cursor…
Access	Whatever they grant	Read-only Google scopes, secrets stay local

How it works

One always-on box between Google Health and your AI

OpenVitals syncs your Google Health data into a local SQLite database and exposes it over the Model Context Protocol. Your AI client connects to that endpoint and calls the tools it needs.

  Google Health API  ──►  OpenVitals server  ──►  SQLite (your machine)
     (your data)          (always-on device)              │
                                                          ▼
                                your AI client  ◄──  POST /mcp  (bearer token)
                            (Claude · ChatGPT)     localhost, or a public URL

🩺

~60 tools over real data

Heart, sleep, activity, glucose, SpO₂, temperature, ECG and body composition — read straight from your Google Health account.

🍎

Nutrition & workouts

Log meals, search foods, scan barcodes, save recipes, build workout plans and search exercises — all through chat.

📈

Summaries that mean something

Daily and weekly rollups, progress tracking, recommendations, and plan-versus-logged comparisons.

🔒

Read-only and local

Google scopes are read-only. Database, tokens and OAuth secrets live on your machine and are gitignored.

🏠

Runs on anything

An old laptop, a Raspberry Pi, a mini-PC. Anything that stays on. This project runs on a laptop-as-homelab-server.

🔌

Any MCP client

Local clients hit localhost directly. Cloud apps reach it through a free ngrok, Cloudflare or Tailscale tunnel.

Quickstart

Running in three steps

You need Node 22+, a Google account with health data, and something that stays powered on. Full walkthrough — including the Google Cloud setup — is in the README.

Create your free Google secrets

Make a free Google Cloud project, enable the Health API, add yourself as a test user, and download an OAuth client JSON. Read-only scopes only.

Clone and run

Install, build the web UI and start the server. A bearer token for /mcp is generated on first run.

Point your AI at it

Add it as an MCP server in Claude Code or Cursor, or paste the URL into a Claude / ChatGPT custom connector.

# 2 — clone and run
$ git clone https://github.com/saiteja007-mv/openvitals.git
$ cd openvitals && npm install
$ npm run build && npm run server   # → http://127.0.0.1:42815

# 3 — connect Claude Code (local, no tunnel needed)
$ claude mcp add --transport http openvitals http://127.0.0.1:42815/mcp \
      --header "Authorization: Bearer $(cat .data/mcp-token.txt)"

Claude Code · Cursor · ClineRun on your machine — no tunnel needed

Claude appLocal bridge, or a custom connector over a tunnel

ChatGPTCustom connector, needs a public HTTPS URL

Anything speaking MCPStreamable HTTP, bearer token or ?token=

Then just ask

Your data, in plain language

"Summarize my sleep this week."

"What's my resting heart rate trend?"

"Log a 5 km run."

"Am I hitting my protein target?"

"Compare my plan to what I actually logged."

Google Health

get_heart · get_sleep · get_activity · get_glucose · get_spo2 · get_temperature · get_body_composition · sync_google_health

Summaries

get_daily_summary · get_weekly_summary · get_progress · get_recommendation · compare_plan_vs_logged

Nutrition

log_meal · search_food · lookup_barcode · get_nutrition_intake · meal recipes

Workouts & habits

log_workout · search_exercises · workout plans · body metrics · habits · reminders · hydration

Your health data, readable by your own AI