garmin-givemydata (100% Working)

It's YOUR data. Take it back. One command. All your Garmin data. Local SQLite + FIT files. Export to CSV, JSON, GPX, TCX. AI-ready via MCP.

Garmin makes it nearly impossible for individuals to access their own health data programmatically. There is no public API. The official "developer program" is restricted to approved businesses only. And in March 2026, Garmin deployed aggressive Cloudflare protections that broke every single community library — garth (deprecated), python-garminconnect (broken auth), and all downstream tools like Home Assistant integrations.

You paid for the hardware. You generated the data with your body. You should be able to access it.

This project gets your data out of Garmin Connect and into a local SQLite database where you own it and AI can analyze it through an MCP server for Claude Code.

50 tables, 10+ years of history, 45 MCP tools for AI analysis. Your data stays on your machine.

The Problem

• No public API: Garmin's Connect Developer Program requires a business application. Individual developers are denied.

• garth is dead: The most popular auth library was deprecated on March 28, 2026 after Garmin changed their SSO flow.

• Bot detection: Garmin deployed aggressive bot detection that blocks all known Python HTTP libraries. Every existing workaround stopped working.

• python-garminconnect is broken: Depends on garth for auth. Issue #332 has 40+ comments from affected users.

• Connect+ paywall: Garmin launched a $7/month subscription and is restricting third-party access to features that compete with it.

Nobody should buy Garmin products until they open their API to the people who paid for the hardware.

Related MCP server: Garmin MCP

Quick Start

Install

# macOS
brew install nrvim/tap/garmin-givemydata

# pip (Linux / Windows / macOS)
pip install garmin-givemydata

# or clone
git clone https://github.com/nrvim/garmin-givemydata.git
cd garmin-givemydata
bash setup.sh          # macOS / Linux
setup.bat              # Windows

Fetch your data

garmin-givemydata                    # fetches all historical data + FIT files

First run prompts for credentials, launches a headless browser, and fetches your full history (~30 min for 10 years). After that, daily syncs take seconds.

Connect AI

Add the MCP server to Claude Code, Claude Desktop, or any MCP client. The config depends on how you installed:

Claude Code (one-liner) — registers the server at user scope so it's available in every project:

# pip / brew install
claude mcp add -s user garmin -- garmin-mcp

# git clone
claude mcp add -s user garmin \
  -e GARMIN_DATA_DIR=/absolute/path/to/garmin-givemydata \
  -- /absolute/path/to/garmin-givemydata/venv/bin/python \
     /absolute/path/to/garmin-givemydata/run_mcp.py

Verify with claude mcp list. Skip the manual JSON below.

Homebrew or pip install — garmin-mcp is already in your PATH:

{
  "mcpServers": {
    "garmin": {
      "command": "garmin-mcp"
    }
  }
}

Git clone — use absolute paths to the venv and set GARMIN_DATA_DIR so the MCP server finds your database:

{
  "mcpServers": {
    "garmin": {
      "command": "/absolute/path/to/garmin-givemydata/venv/bin/python",
      "args": ["/absolute/path/to/garmin-givemydata/run_mcp.py"],
      "cwd": "/absolute/path/to/garmin-givemydata",
      "env": {
        "GARMIN_DATA_DIR": "/absolute/path/to/garmin-givemydata"
      }
    }
  }
}

Note: GARMIN_DATA_DIR tells the MCP server where garmin.db lives. Without it, the server falls back to ~/.garmin-givemydata/ which may not be where your data is if you cloned to a custom location.

Save this as:

• Claude Code: .mcp.json in your project root, or ~/.claude/settings.json under mcpServers for global access

• Claude Desktop: ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows)

Restart your client and run /mcp to approve the server. Then ask:

"How was my sleep this week?"
"Am I overtraining? Check my HRV and recovery"
"Compare my fitness this month vs last month"
"Give me a full sports medicine health check"

Windows (git clone)

{
  "mcpServers": {
    "garmin": {
      "command": "C:\\Users\\jane\\code\\garmin-givemydata\\venv\\Scripts\\python.exe",
      "args": ["C:\\Users\\jane\\code\\garmin-givemydata\\run_mcp.py"],
      "cwd": "C:\\Users\\jane\\code\\garmin-givemydata",
      "env": {
        "GARMIN_DATA_DIR": "C:\\Users\\jane\\code\\garmin-givemydata"
      }
    }
  }
}

Other MCP clients (OpenClaw, Cline, Cursor)

Any client supporting MCP stdio transport can connect. For pip/brew: garmin-mcp. For git clone:

/path/to/garmin-givemydata/venv/bin/python /path/to/garmin-givemydata/run_mcp.py

Important: Paths must be absolute — relative paths will not work.

What You Get

• ALL your data in one command — 50 tables, activities with splits/weather/HR zones/GPS trackpoints, original FIT files

• 10+ years of history fetched automatically, smart incremental sync after that

• 45 MCP tools for AI analysis — not just raw data, but tools with clinical context, anomaly detection, and professional training metrics

• Export to anything — CSV, JSON, GPX, TCX from your local database

• Your data stays local — nothing is sent anywhere

MCP Tools (45)

The MCP server gives AI assistants deep access to your health data. Every tool returns data + context — not just numbers, but trend direction, anomaly flags, clinical thresholds, and goal attainment.

Usage

Fetch (from Garmin Connect)

garmin-givemydata                              # smart sync (all data + FIT files)
garmin-givemydata --full                       # force full historical re-fetch
garmin-givemydata --days 90                    # fetch last 90 days
garmin-givemydata --since 2025-01-01           # fetch from specific date
garmin-givemydata --profile health             # health metrics only
garmin-givemydata --profile activities         # activities + FIT files only
garmin-givemydata --profile sleep              # sleep data only
garmin-givemydata --no-files                   # skip FIT file downloads
garmin-givemydata                              # parses trackpoints for newly downloaded FIT files by default
garmin-givemydata --no-trackpoints             # skip trackpoint parsing during sync
garmin-givemydata --rebuild-trackpoints        # full rebuild: reparse all downloaded FIT files
garmin-givemydata --status                     # check database contents

FIT file download only

garmin-givemydata --fit-only --latest          # latest FIT file
garmin-givemydata --fit-only --date 2026-03-30 # specific date
garmin-givemydata --fit-only --days 7          # last 7 days
garmin-givemydata --fit-only                   # all FIT files

Export (from local database)

garmin-givemydata --export ./output            # CSV + JSON
garmin-givemydata --export-gpx ./gpx           # GPX (for Strava, Komoot)
garmin-givemydata --export-tcx ./tcx           # TCX (for TrainingPeaks)

Browser engine

Uses SeleniumBase UC mode (undetected Chrome) for Cloudflare bypass. Requires Google Chrome installed.

Note: cf_clearance is bound to your IP address. If your IP changes, the session expires regardless of profile state. For unattended use, run on a machine with a stable egress IP.

garmin.db              # SQLite database (all health + activity data)
browser_profile/       # Browser session (persists ~1 year)
.env                   # Garmin credentials
fit/                   # Original FIT files (lossless)

Architecture

garmin-givemydata/
├── garmin_givemydata.py       # Main entry: smart sync (full or incremental)
├── garmin_client/              # SeleniumBase-based Garmin Connect client
│   ├── client.py               #   GarminClient (login, fetch, export)
│   └── endpoints.py            #   API endpoint definitions
├── garmin_mcp/                 # MCP server + database layer
│   ├── db.py                   #   SQLite schema (50 tables), upsert helpers
│   ├── server.py               #   FastMCP server with 45 tools
│   ├── export.py               #   CSV, JSON, GPX, TCX export
│   ├── import_json.py          #   JSON → SQLite bulk import
│   └── sync.py                 #   Incremental sync engine
├── run_mcp.py                  # MCP server entry point
├── garmin.db                   # Your health data (SQLite, gitignored)
├── fit/                        # Your activity files (FIT/ZIP, gitignored)
├── browser_profile/            # Browser session (gitignored)
└── pyproject.toml

Data flow:

Garmin Connect ──→ SQLite (health + activity metrics)
                └─→ fit/ (original FIT files, lossless)

SQLite ──→ MCP server (AI queries via 45 tools)
       ├─→ CSV/JSON (--export)
       └─→ GPX/TCX (--export-gpx, --export-tcx)

Platform Support

brew install python@3.12
cd garmin-givemydata
python3.12 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
# Chrome must be installed (https://www.google.com/chrome/)
cp .env.example .env

sudo apt update && sudo apt install python3.12 python3.12-venv
cd garmin-givemydata
python3.12 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
# Chrome must be installed (https://www.google.com/chrome/)
cp .env.example .env

sudo dnf install python3.12
cd garmin-givemydata
python3.12 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
# Chrome must be installed (https://www.google.com/chrome/)
cp .env.example .env

Install Python 3.10+ from python.org — check "Add to PATH".

cd garmin-givemydata
python -m venv venv
venv\Scripts\Activate.ps1
pip install -r requirements.txt
# Chrome must be installed (https://www.google.com/chrome/)
copy .env.example .env

If PowerShell blocks the activate script: Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser

Troubleshooting

"Login failed": Delete the browser profile and run again. For pip/brew: rm -rf ~/.garmin-givemydata/browser_profile. For git clone: rm -rf browser_profile/.

Script crashed: Don't close the Chrome window manually. The tool handles shutdown — killing Chrome mid-run corrupts the profile. Run again (stale locks are auto-cleaned).

"Python not found": Make sure Python 3.10+ is on your PATH. macOS: brew install python@3.12. Ubuntu: sudo apt install python3.12 python3.12-venv.

ensurepip is not available / venv creation fails: Your system Python is missing the venv module. Install the matching package: sudo apt install python3.X-venv (replace X with your minor version — python3 --version to check). On Ubuntu with Python 3.14, that's python3.14-venv. The setup.sh script auto-detects Python but can't auto-install the venv package.

403 or session errors: Session expired (often from IP change — cf_clearance is IP-bound). Delete the browser profile and re-login.

Chrome doesn't open (Linux/SSH): Install Xvfb: sudo apt install xvfb. The tool auto-spawns Xvfb at 1920x1080 when no display is available, or use xvfb-run -a garmin-givemydata.

Session rot over SSH: Run under systemd-run --user --scope or tmux/screen so SSH disconnect doesn't SIGHUP the process. The tool installs signal handlers but a detached session is more reliable.

MCP server "failed to connect": Paths in .mcp.json must be absolute. Test: <path>/venv/bin/python <path>/run_mcp.py

Empty data for some metrics: Training readiness, HRV, body battery, endurance score, hill score, and race predictions require compatible devices (Fenix 7+, Forerunner 265+, Venu 3+, etc.).

Windows: "execution of scripts is disabled": Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser

Contributing

• New endpoints: Garmin has hundreds of internal APIs. Discover new ones via browser dev tools and add them to endpoints.py.

• More MCP tools: The server has 45 tools including CTL/ATL/TSB, recovery signatures, and period comparison. Ideas: injury risk prediction, sleep optimization, race readiness scoring, overtraining detection.

• MCP client integrations: Test with OpenClaw, Cline, Continue, Cursor, or other clients.

• Other platforms: ARM (Raspberry Pi), Docker, etc.

• Data visualization: Dashboards, charts, reports from SQLite.

• Export formats: Parquet, or other formats for pandas, R.

• Other wearables: The architecture (browser auth + SQLite + MCP) could work for COROS, Samsung, Amazfit/Zepp.

• Testing: Tests, CI, more platform support.

Open an issue or submit a PR.

Support This Project

• Star this repository — helps others discover the project

• Report issues — improve stability and compatibility

• Spread the word — share with other Garmin users

• Contribute code — new endpoints, MCP tools, export formats, bug fixes

• File a GDPR/CCPA request with Garmin — the more users who formally request data portability, the harder it is to ignore

Acknowledgments

• garth by @matin — 350k+ monthly PyPI downloads before deprecation. The auth library that powered the Garmin Python ecosystem.

• python-garminconnect by @cyberjunky — 2,000+ stars, 127+ API methods. The endpoint catalog was informed by their work.

• GarminDB by @tcgoetz — 3,000+ stars. Pioneered SQLite-first Garmin data storage.

• garmin-connect-export — Reliable activity export (GPX, TCX, FIT, JSON).

• garmin-data-export — Garmin data into LLM-readable text files.

• garmy — AI-first Garmin data access with MCP support.

EU/EEA — GDPR

Article 20 — Right to Data Portability grants the right to receive personal data in a "structured, commonly used and machine-readable format." This covers raw sensor data (heart rate, sleep, steps, SpO2). Derived metrics (training readiness, fitness age) are covered under Article 15 — Right of Access.

United States — CCPA

Section 1798.100(d) requires businesses to provide personal information in a "readily useable format that allows the consumer to transmit this information from one entity to another entity without hindrance."

EU Data Act (Since September 2025)

Article 3 explicitly requires IoT manufacturers to make data available "without undue delay, free of charge, and, where applicable, continuously and in real-time." A Garmin watch is an IoT device. The data it generates belongs to the user.

If Garmin provided an open API, this tool would not need to exist.

Disclaimer

This tool is unofficial and not affiliated with Garmin. It accesses Garmin Connect using your own credentials to retrieve your own data. Use may violate Garmin's Terms of Service. You assume all risk and responsibility. See the full disclaimer for details.

License

AGPL-3.0 — Free to use, modify, and distribute. If you build a service using this code, you must keep it open source.