fitness-mcp

A personal remote MCP (Model Context Protocol) server that lets Claude read your Hevy workout data and MacroFactor nutrition data directly in conversation. Deployed on Vercel's free Hobby tier.

Status

• Hevy: implemented (read-only: recent workouts, workout detail, body measurements). Uses the official Hevy REST API directly.

• MacroFactor: implemented, but indirectly. MacroFactor's backend enabled Firebase App Check enforcement in May 2026, which blocks all third-party API access — there is no legitimate way to pull MacroFactor data live. Instead, this server reads the Google Sheets files produced by MacroFactor's manual "Granular Export" feature, which the user saves into a "Health data" Drive folder. Every MacroFactor tool call lazily syncs any newly exported files first, so a fresh manual export becomes visible to Claude on the very next question — no separate trigger/webhook needed.

Related MCP server: hevy-mcp

Tools exposed

All tools are read-only.

How the MacroFactor sync works

MacroFactor has no usable API, so data instead flows: MacroFactor app → manual "Granular Export" → Google Drive folder "Health data" → this server reads it on demand.

• Each export creates a new MacroFactor-<timestamp> Google Sheet in that folder (never overwritten). Export date ranges can overlap or have gaps — the sync does not assume the latest file has full history.

• On every MacroFactor tool call, the server checks the Drive folder for export files it hasn't processed yet, merges any new ones into a small JSON store (Health data/_store/daily-summary.json), and recomputes precomputed monthly/yearly rollups for just the affected periods.

• If nothing new was exported since the last call, this is a fast no-write no-op — repeated questions ("how's my year going?") don't re-scan or re-download anything.

• When two exports cover the same date, the more recently exported one wins (compared by the timestamp encoded in the file name), regardless of Drive's file ordering.

Setting up Google Drive access (one-time)

1. Create or select a Google Cloud project.

2. Enable the Google Drive API and Google Sheets API for it.

3. Create a Service Account (IAM & Admin → Service Accounts). No IAM roles needed.

4. Create and download a JSON key for the service account.

5. Base64-encode it: base64 -w0 key.json

6. Set the result as the GOOGLE_SERVICE_ACCOUNT_KEY_BASE64 env var (see below).

7. In Google Drive, share the "Health data" folder with the service account's email (the client_email field in the JSON key) as Editor — the app creates and writes to a _store subfolder inside it.

8. No domain-wide delegation needed — this is a personal Gmail account, a plain folder share is enough.

Local development

npm install
cp .env.example .env.local   # fill in real values
vercel dev

Smoke test (replace $MCP_BEARER_TOKEN):

curl -X POST http://localhost:3000/api/mcp \
  -H "Authorization: Bearer $MCP_BEARER_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Should return the 6 tools above. A request with a missing/wrong token should get 401.

Testing

Three layers, all run in CI (.github/workflows/ci.yml) on every push/PR — none require real Hevy/Google secrets, so they work the same in a public repo:

npm run test        # unit + integration (vitest) — pure logic, plus lib/googleDrive.ts
                     # mocked with an in-memory fake Drive, plus the real Next.js
                     # route handler exercised with fetch/Drive mocked
npm run build
npm run test:e2e     # starts a real `next start` server and hits it over real HTTP
                      # (node's built-in test runner, no extra dependency)

• Unit (lib/*.test.ts): date parsing, MacroFactor tab parsing, monthly/yearly rollup math, bearer-token verification.

• Integration (test/integration/*.test.ts): the MacroFactor sync algorithm (multi-export merge, overlap resolution, incremental rollups) against test/fixtures/fakeGoogleDrive.ts; and the actual app/api/mcp/route.ts handler wired to real lib/auth.ts/lib/hevy.ts/lib/macrofactorStore.ts, with only fetch and Drive mocked.

• E2E (test/e2e/*.e2e.test.mjs): boots the production build and asserts over real HTTP — health check, 401 on bad/missing auth, tools/list returns all 6 tools. Doesn't exercise real Hevy/MacroFactor data (CI has no real credentials by design).

Environment variables

Set these in the Vercel project's Environment Variables (Production + Preview). Never commit real values — .env.example only documents the names.

Deploy

1. vercel link

2. vercel env add HEVY_API_KEY / vercel env add MCP_BEARER_TOKEN / vercel env add GOOGLE_SERVICE_ACCOUNT_KEY_BASE64 (repeat for each environment you use)

3. Connect this GitHub repo in the Vercel dashboard for auto-deploy on push to main, or run vercel --prod manually.

4. Note the deployed URL, e.g. https://fitness-mcp.vercel.app/api/mcp.

Connect to Claude

Custom connectors can only be added from claude.ai (web) or the desktop app — not from the mobile app. Once added there, they're usable from mobile automatically.

1. On claude.ai: Settings → Connectors → Add custom connector.

2. Name: Fitness Data. URL: https://fitness-mcp.vercel.app/api/mcp.

3. Under "Request headers", add Authorization: Bearer <MCP_BEARER_TOKEN> (the same value set in Vercel).

4. Save. Claude should list the 6 tools above.

Try asking: "直近のワークアウトを教えて" (tell me about my recent workouts) or "今月の平均カロリーは?" (what's my average calorie intake this month?).