google-health-worker-mcp
A personal MCP server on Cloudflare Workers that connects Claude to Google Health API v4, enabling reading of health data like activity, sleep, heart rate, SpO₂, HRV, and nutrition.
README
google-health-worker-mcp
Personal remote MCP server on Cloudflare Workers that connects Claude (iPhone / desktop / web via Custom Connector) to Google Health API v4 — read activity, sleep, heart rate, SpO₂, HRV, respiratory rate, skin temperature, weight and nutrition.
All 16
HealthProviderread methods run on Google Health v4 (GoogleHealthProvider). This edition exposes the read surface; the write/delete methods are present in the interface but not implemented.
Based on tachibanayu24/fitbit-googlehealth-mcp (MIT). Upstream docs preserved under docs/upstream-*.
Why
The legacy Fitbit Web API shuts down in September 2026. Google Health API v4 is its successor. This project keeps the upstream's Workers / Hono / MCP architecture — the only form factor usable from the Claude mobile app — and swaps the data source to v4.
Capabilities
| Domain | Methods | Status |
|---|---|---|
| Profile · devices | getProfile, listDevices |
✅ live |
| Activity | getActivityTimeSeries (steps/distance/calories/floors), getDailySummary |
✅ live |
| Exercise | getExerciseList |
✅ live (mapping unverified — no exercise data on account) |
| Sleep | getSleep, getSleepRange (stage mapping) |
✅ live |
| Heart | getHeartRateRange (resting), getHeartRateIntraday (downsampled) |
✅ live |
| Metrics | getHRV, getSpO2, getRespiratoryRate, getSkinTemperature, getCardioFitness |
✅ live |
| Body · food | getBodyLog (weight), getFoodLog (nutrition) |
✅ live |
The HealthProvider interface also declares write/delete methods (logFood, logWeight, …); in this edition they throw "not implemented".
Skin temperature is confirmed absolute °C + baseline; the mapper reports the nightly-relative delta. Reconcile calls omit filter (v4 filter members are type-specific and mostly 400); range narrowing is done client-side on the civil date.
Setup
One-time bootstrap.
Google Cloud (console):
- Create a project → APIs & Services → enable Health API.
- OAuth consent screen (User type External): add the read scopes below, add your own Gmail as a test user, then Publish the app to "In production". (Testing-mode refresh tokens expire after 7 days, which breaks an unattended Worker.)
- Credentials → create an OAuth client (Web application), Authorized redirect URI
http://127.0.0.1:8787/oauth/callback. Note the client id / secret.
Read scopes (prefix https://www.googleapis.com/auth/):
googlehealth.profile.readonly
googlehealth.settings.readonly
googlehealth.activity_and_fitness.readonly
googlehealth.health_metrics_and_measurements.readonly
googlehealth.sleep.readonly
googlehealth.nutrition.readonly
Do not add
include_granted_scopesto the authorization URL — mixing legacy scopes can cause403s on reads. Useaccess_type=offline&prompt=consentto get a refresh token.
This repo:
pnpm install # you run this — deps are never auto-installed
cp wrangler.toml.example wrangler.toml
pnpm wrangler kv namespace create TOKENS # paste ids into wrangler.toml
pnpm wrangler kv namespace create CACHE
export GOOGLE_CLIENT_ID=... # the Web OAuth client from above
export GOOGLE_CLIENT_SECRET=...
pnpm run setup:google # browser consent -> prints wrangler secret/kv commands
pnpm deploy # then add the Custom Connector in claude.ai:
# https://<worker>.workers.dev/mcp/<MCP_SHARED_SECRET>
MCP_SHARED_SECRET must be hex (openssl rand -hex 32) — it is embedded in the URL path, so base64's / would break routing. Secrets live in Cloudflare, tokens in KV — never in this repo.
Architecture
Claude (mobile / desktop / web)
│ Streamable HTTP /mcp/<SECRET>
▼
Cloudflare Workers
├─ guard (shared secret + Anthropic CIDR allowlist)
├─ @hono/mcp (Streamable HTTP transport)
└─ McpServer → tools/* → HealthProvider
└─ GoogleHealthProvider → health.googleapis.com/v4
KV: TOKENS (Google OAuth refresh/access) · CACHE (read cache)
Two auth layers: ① Claude → Worker = shared secret in the URL path + Anthropic CIDR allowlist; ② Worker → Google = KV refresh token, auto-refreshed. The tools layer is provider-agnostic (HealthProvider); swapping Fitbit → Google was one line in src/server.ts.
Development
pnpm typecheck # tsc --noEmit
pnpm test # vitest (113 tests)
pnpm lint # biome
pnpm deploy # wrangler deploy
Provider field mappings are covered by fixture tests using shapes probed from the live v4 API. Deployment: connect the repo in Cloudflare Workers Builds (dashboard → Workers → Builds → connect GitHub) so push main runs install + test + wrangler deploy automatically.
Documents
| Doc | What |
|---|---|
docs/upstream-*.md |
Upstream README / research / journal (Fitbit-era reference) |
License
MIT — see LICENSE (upstream copyright preserved).
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.