mcpField
G
Garmin Connect Personal API

Garmin Connect Personal API

Local
github.comHealth0Updated 3d ago

About Garmin Connect Personal API

Garmin Connect Personal API MCP is a read-only server that lets AI assistants access personal Garmin Connect training, recovery and health data. It can run locally or on the user's Cloudflare account, and provides detailed activity, sleep and wellness records.

Key features

  1. 01Read-only Garmin Connect data access
  2. 02Activity details, streams and laps
  3. 03Sleep and HRV records
  4. 04Daily recovery and health summaries
  5. 05Weight, hydration and food logs
  6. 06Wellness and cycle data

Use cases

  1. 01Review training activities and workout metrics
  2. 02Track sleep and HRV across a date range
  3. 03Inspect daily recovery and heart-rate data
  4. 04Review hydration, nutrition or cycle records

MCP support

Exposes read-only Garmin Connect Personal API data through MCP tools. It can run locally or be hosted on Cloudflare for remote MCP clients.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"hannta-garmin-connect-personal-api": {
"command": "npx",
"args": [
"-y",
"@hannta/garmin-connect-personal-api-mcp"
],
"env": {
"GARMIN_PERSONAL_API_KEY": "<GARMIN_PERSONAL_API_KEY>"
}
}
}
}
Run commandnpx -y @hannta/garmin-connect-personal-api-mcp
Environment variables to set
GARMIN_PERSONAL_API_KEYSecretGarmin Connect Personal API key (optional: the auth command saves one instead)

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
npm@hannta/garmin-connect-personal-api-mcp1.1.1

Tools

27

Saved tool list · updated 1w ago

ParameterTypeRequiredDescription
datestringNoCalendar date, YYYY-MM-DD. Default: today in the user's time zone.

Details

From the project README
Watch with a heart-rate line

Garmin Connect Personal API MCP

Not affiliated with or endorsed by Garmin. Garmin and Garmin Connect are trademarks of Garmin Ltd.

A read-only MCP server that gives AI assistants your own Garmin Connect training, recovery and health data through the official Garmin Connect Personal API. Run it on your own computer, or host it on your own Cloudflare account to use it from anywhere.

AI agents: to set this up for a user, follow docs/AGENT-SETUP.md.

Setup

You need a Garmin Connect Personal API key, which you create in Garmin Connect on the web.

Run locally

The quickest way, for Claude Code, Claude Desktop, Cursor, VS Code and other clients that run local servers. Needs Node.js 22+.

Paste this into your coding agent:

Connect @hannta/garmin-connect-personal-api-mcp

Or do it yourself: docs/LOCAL.md.

Host on Cloudflare

For Claude on the web and on mobile, and any other client that connects to remote servers with OAuth. The free plan is enough, apart from one known limitation.

Then connect your MCP client, such as Claude, ChatGPT, Cursor or VS Code.

Tools

ToolWhat it returns
garmin_get_dayOne day at a glance: sleep, HRV, Body Battery, stress, resting HR, intensity minutes, activities
garmin_get_daysThe same over a date range, one row per day, with daily activity totals and training load
garmin_search_activitiesActivities in a date range with Garmin IDs and fastest splits, or totals per week
garmin_get_activityOne activity in depth: pace, HR zones, running dynamics, power, training effect, interval totals, strength sets per exercise
garmin_get_activity_streamsOne activity's time series (HR, pace, grade-adjusted pace, cadence, power, elevation, …)
garmin_get_activity_lapsOne row per lap and interval rep, from the activity's FIT file
garmin_get_sleepOne night's sleep in detail: score components with optimal ranges, sleep need and why it changed, naps
garmin_get_sleep_summariesSleep per night over a date range
garmin_get_hrvOne night's HRV with baseline, status and overnight readings
garmin_get_hrv_summariesHRV per night over a date range
garmin_get_daily_summariesResting HR, Body Battery, stress, intensity minutes, steps per day over a date range
garmin_get_weightWeigh-ins and body composition over a date range
garmin_get_profileProfile settings and heart-rate zones
garmin_get_body_batteryBody Battery per day: charged, drained and the sleep / activity events behind it
garmin_get_stressStress per day: average level and minutes at rest / low / medium / high
garmin_get_stepsSteps per day with goal and distance
garmin_get_heart_rateOne day's resting / min / max heart rate, optionally averaged per 15 minutes
garmin_get_spo2One day's blood oxygen (Pulse Ox), optionally hourly with altitude
garmin_get_fitness_ageFitness age and its components (resting HR, BMI, vigorous days / minutes)
garmin_get_heart_rate_zonesHeart-rate zones per sport
garmin_get_weight_goalsWeight goals in effect (target kg)
garmin_get_activity_workoutsThe structured workout an activity followed: steps, targets (HR, pace, power) and repeats
garmin_get_hydrationHydration per day: intake, goal, sweat loss (ml)
garmin_get_food_logOne day's food log: calories and macros eaten vs. goals, per meal and food
garmin_get_wellness_activitiesWellness sessions (e.g. Health Snapshot, breathwork) with HR, stress, HRV, SpO2
garmin_get_menstrual_cycleCycle day, period and fertile window, predictions and cycle analysis
garmin_get_menstrual_dayOne day's logged flow, symptoms and moods, with that day's cycle summary
Good to know
  • Range tools default to the last 7 days. If a range is too long for one Garmin request, the error says the maximum and the assistant splits it into several calls.
  • Output is metric, with the unit in every field name. Dates are YYYY-MM-DD in your time zone; sleep and HRV are keyed by the wake-up date.
  • Location, sexual activity and sex drive data are never returned.
  • Known limitation: garmin_get_activity_laps can fail on long activities on the Cloudflare free plan. See Known limitations.

Development

Requires Node.js 24+.

npm install
cp .dev.vars.example .dev.vars   # then add GARMIN_API_KEY and OWNER_PASSPHRASE
npm run check                     # typecheck, lint, format check, tests
npm run today                     # print today's recovery data using your key
npm run dev                       # MCP server at http://localhost:8787/mcp
npm run build                     # local server in dist/cli.mjs

To try the local server, connect the MCP Inspector to it and sign in with OWNER_PASSPHRASE.

Architecture and conventions: docs/DEVELOPMENT.md.

License

MIT

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.