Skip to main content
Glama
Garmin Connect MCP

Garmin Connect MCP

by LePomas

Garmin Connect MCP

An MCP server that exposes your personal Garmin Connect data to MCP-capable clients — Claude Desktop, Claude Code, Codex, and Gemini CLI.

The Garmin work is done by the python-garminconnect library; this repo is the thin MCP layer on top: tool definitions, a stdio server, and privacy guardrails that keep raw health payloads out of durable text. Because python-garminconnect relies on Garmin Connect's unofficial client behavior, it can break if Garmin changes login or endpoint behavior.

Setup

cd /path/to/garminconnect-mcp
python3 -m venv .venv
.venv/bin/python -m pip install -e .
cp .env.example .env

Edit .env with GARMIN_EMAIL and GARMIN_PASSWORD.

Run once from a terminal to create or refresh saved Garmin tokens:

.venv/bin/garminconnect-mcp login

If Garmin asks for MFA, the login command prompts for the code in the terminal. For non-interactive use, set GARMIN_MFA_CODE temporarily.

The MCP server runs over stdio. Configure Codex with this command:

{
  "mcpServers": {
    "garmin": {
      "command": "/path/to/garminconnect-mcp/.venv/bin/garminconnect-mcp"
    }
  }
}

Claude config files are included too:

  • .mcp.json for project-scoped Claude Code use.

  • docs/claude_desktop_config.json for Claude Desktop.

  • docs/claude.md for setup and verification steps.

Gemini CLI config files are included too:

  • .gemini/settings.json for project-scoped Gemini CLI use.

  • docs/gemini_settings.json for user-scoped Gemini CLI config.

  • docs/gemini.md for setup and verification steps.

Related MCP server: Garmin MCP Server

Tools

Connection-only tools:

  • garmin_connection_status

  • garmin_ping

Raw private Garmin data tools:

  • garmin_profile

  • garmin_daily_stats

  • garmin_heart_rate

  • garmin_sleep

  • garmin_hrv

  • garmin_body_battery

  • garmin_stress

  • garmin_recent_activities (compact summary by default; compact=false for raw)

  • garmin_activity

  • garmin_activity_laps (per-lap summary with workout step index + compliance)

  • garmin_body_composition (weight/muscle/bone kg, fat %, water %, BMI)

Summarized workout tools:

  • garmin_workouts

  • garmin_scheduled_workouts

  • garmin_schedule_workout

  • garmin_create_scheduled_workout

  • garmin_unschedule_workout

Dates use YYYY-MM-DD. If omitted, tools default to today.

Use garmin_connection_status or garmin_ping for smoke tests. They validate login without returning profile, health, or account data.

This is a personal local MCP server, and the raw tools intentionally return full Garmin payloads. Avoid pasting those raw responses into docs, examples, issues, or other durable text unless you have sanitized them.

Workout tools return summarized fields instead of raw Garmin payloads. To create and schedule a new workout, pass Garmin Connect workout JSON to garmin_create_scheduled_workout; to schedule an existing template, use garmin_schedule_workout with its workout ID.

Development

scripts/check-private-output.sh
.venv/bin/python -m pytest
.venv/bin/python -m ruff check .
.venv/bin/python -m ruff format --check .
.venv/bin/python -m compileall src

Available Tools

16 tools
garmin_activityC

Get raw private Garmin activity details by activity ID.

Inputs: activity_id (required). Returns: object

garmin_body_batteryA

Get raw private Body Battery data for a date in YYYY-MM-DD format.

Inputs: day. Returns: result

garmin_connection_statusA

Check Garmin Connect login without returning personal data.

Inputs: no named parameters. Returns: object

garmin_create_scheduled_workoutB

Upload a Garmin workout JSON payload, then schedule it on a YYYY-MM-DD date.

Inputs: day (required), workout_json (required). Returns: object

garmin_daily_statsB

Get raw private daily Garmin stats for a date in YYYY-MM-DD format.

Inputs: day. Returns: object

garmin_heart_rateA

Get raw private heart-rate data for a date in YYYY-MM-DD format.

Inputs: day. Returns: object

garmin_hrvB

Get raw private HRV data for a date in YYYY-MM-DD format.

Inputs: day. Returns: object

garmin_pingA

Check Garmin Connect login without returning personal data.

Inputs: no named parameters. Returns: object

garmin_profileA

Get the raw private Garmin Connect user profile.

Inputs: no named parameters. Returns: object

garmin_recent_activitiesC

List raw private recent Garmin activities.

Inputs: limit, start. Returns: result

garmin_scheduled_workoutsA

List scheduled Garmin workouts for a calendar month using 1-12 months.

Inputs: year (required), month (required). Returns: object

garmin_schedule_workoutB

Schedule an existing Garmin workout template on a YYYY-MM-DD date.

Inputs: day (required), workout_id (required). Returns: object

garmin_sleepA

Get raw private sleep data for a date in YYYY-MM-DD format.

Inputs: day. Returns: object

garmin_stressB

Get raw private stress data for a date in YYYY-MM-DD format.

Inputs: day. Returns: object

garmin_unschedule_workoutB

Remove a Garmin workout from the calendar without deleting the template.

Inputs: scheduled_workout_id (required). Returns: object

garmin_workoutsC

List saved Garmin workout templates with summarized fields.

Inputs: limit, start. Returns: object

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    B3.4/5.0

    Scored across 16 tools

    Disambiguation4/5

    Most tools target distinct data types or actions, but there is direct overlap between garmin_connection_status and garmin_ping, both described as checking login status without returning personal data. Also, workout scheduling tools (garmin_create_scheduled_workout vs garmin_schedule_workout) may cause slight confusion, though descriptions clarify differences.

    Naming Consistency5/5

    All tools follow a consistent 'garmin_' prefix followed by snake_case nouns or verb-noun phrases (e.g., garmin_activity, garmin_create_scheduled_workout). The naming pattern is uniform and predictable.

    Tool Count5/5

    With 16 tools covering diverse health and fitness data (activities, body battery, heart rate, sleep, stress, workouts, profile, etc.), the count is well-scoped for a personal data access server. Each tool serves a clear purpose without being excessive.

    Completeness4/5

    The tool set covers core query and some mutation operations for Garmin data. However, gaps exist: no tool to delete workout templates (only unschedule), no activity deletion, and no date-range filtering for activities beyond recent. Overall, minor missing features for a comprehensive Garmin management surface.

    Maintenance

    ActivityMaintained
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers