Google Health MCP Server

An MCP server that connects any MCP-compatible AI client to the Google Health API, giving you access to your health and fitness data directly in AI conversations.

Runs as a Cloudflare Worker — deploy your own instance in ~10 minutes.

Compatible with Claude, Cursor, Windsurf, Zed, VS Code (Copilot), Continue.dev, and any other MCP-compatible client.

What you can ask

• "How many steps did I take last week?"

• "Show me my sleep data for the past month"

• "What was my resting heart rate trend in May?"

• "Summarize my exercise sessions this year"

• "How does my SpO2 look over the last 30 days?"

Related MCP server: Google Health MCP

Available tools

Prerequisites

• Cloudflare account (free tier is sufficient)

• Google Cloud project with the Health API enabled

• Node.js 18+

Setup

1. Clone and install

git clone https://github.com/YOUR_USERNAME/google-health-mcp.git
cd google-health-mcp
npm install

2. Enable the Google Health API

1. Go to Google Cloud Console → APIs & Services → Enable APIs

2. Search for and enable Google Health API

3. Go to Credentials → Create OAuth 2.0 Client ID (type: Web application)

4. Add http://127.0.0.1:8765/callback as an authorised redirect URI

5. Note your Client ID and Client Secret

3. Create a Cloudflare KV namespace

npm run kv:create

Copy the returned id and paste it into wrangler.toml under [[kv_namespaces]].

4. Set secrets in Cloudflare

# A long random string — becomes part of your private connector URL
wrangler secret put MCP_SHARED_SECRET

# From your Google Cloud OAuth client
wrangler secret put GOOGLE_CLIENT_ID
wrangler secret put GOOGLE_CLIENT_SECRET

5. Get your Google refresh token

GOOGLE_CLIENT_ID=your-id GOOGLE_CLIENT_SECRET=your-secret npm run token:setup

This opens a browser for Google consent. After approving, copy the printed command and run it:

wrangler kv key put --binding=HEALTH_TOKENS google_refresh_token "your-refresh-token"

6. Deploy

npm run deploy

Your MCP server is now live at:

https://google-health-mcp.<your-workers-subdomain>.workers.dev/mcp/<MCP_SHARED_SECRET>

7. Connect your MCP client

Add the URL above as a custom MCP server in your AI client. Some examples:

• Claude.ai → Customize → Integrations → Add integration URL

• Cursor → Settings → MCP → Add server URL

• Windsurf → Settings → MCP Servers → Add

• VS Code (Copilot) → .vscode/mcp.json → add server entry

• Continue.dev → config.json → mcpServers array

Local development

cp .dev.vars.example .dev.vars
# Fill in .dev.vars with your credentials
npm run dev

Data types

The get_raw_data_points tool accepts any identifier from the Google Health data types reference, e.g. body-fat, vo2-max, blood-glucose.

Security

• The MCP endpoint is only accessible via a secret URL — treat it like a password

• This server is read-only: it never writes data back to Google Health

• Your refresh token is stored in Cloudflare KV, encrypted at rest

Staying up to date

To pull the latest changes and redeploy your instance:

git pull origin main
npm install   # only needed if dependencies changed
npm run deploy

No re-setup of secrets or KV is needed — those persist across deployments.

If new tools aren't showing up in your AI client after a deployment, disconnect and reconnect the integration — most MCP clients cache the tool list from when you first connected.

If a release changes the Google OAuth scopes (check the CHANGELOG), you'll need to re-run npm run token:setup to get a new refresh token with the updated permissions.

Changelog

See CHANGELOG.md for a full history of releases and what changed in each version.

License

MIT