garmin-mcp
Provides read access to Garmin Connect health and training data (sleep, heart rate, stress, HRV, activities) and supports writes like creating workouts, editing activities, and exporting files.
README
garmin-mcp
An MCP server that gives Claude read access to your Garmin Connect data — sleep, heart rate, stress, training load, HRV, activities — plus a small number of writes (create a structured workout, edit an activity, export a file).
TypeScript, plain node. No uvx, no Python, no Docker.
Setup
npm install
npm run auth # one-time Garmin login; caches tokens
npm run build
npm run auth prompts for your Garmin email and password, exchanges them for
OAuth tokens, and writes ~/.garmin-mcp/tokens.json with 0600 permissions.
Your password is never stored. The tokens last roughly 30 days, after which you
re-run the command.
Accounts with two-factor authentication are not supported — the underlying library cannot answer an MFA challenge.
Connecting Claude Desktop
Find claude_desktop_config.json:
| OS | Path |
|---|---|
| Windows | %APPDATA%\Claude\claude_desktop_config.json |
| macOS | ~/Library/Application Support/Claude/claude_desktop_config.json |
You can also reach it from Claude Desktop: Settings → Developer → Edit Config. Create the file if it doesn't exist.
Option A — cached token file (recommended)
Nothing secret goes in the config; the server reads ~/.garmin-mcp/tokens.json.
{
"mcpServers": {
"garmin": {
"command": "node",
"args": ["C:\\Users\\abhim\\Documents\\garmin-mcp\\dist\\stdio.js"]
}
}
}
On macOS the path is plain, with no escaping:
{
"mcpServers": {
"garmin": {
"command": "node",
"args": ["/Users/you/garmin-mcp/dist/stdio.js"]
}
}
}
Option B — token string in the environment
For machines where you'd rather not leave a token file, or where the server runs
somewhere without your home directory. Use the GARMIN_TOKENS_BASE64 value
printed by npm run auth.
{
"mcpServers": {
"garmin": {
"command": "node",
"args": ["C:\\Users\\abhim\\Documents\\garmin-mcp\\dist\\stdio.js"],
"env": {
"GARMIN_TOKENS_BASE64": "eyJvYXV0aDEiOnsi..."
}
}
}
}
The environment variable wins when both are present, so Option B overrides a cached file rather than racing it. Note that anything in this block sits in a plaintext config file — the token grants full account access until it expires, so Option A is the better default.
Restart Claude Desktop after editing. The tools appear under the plug icon.
If it doesn't connect
- Use an absolute path. Claude Desktop does not run the server from your project directory, so a relative path won't resolve.
- Run
npm run buildfirst. The config points atdist/stdio.js, not the TypeScript source. - Check
nodeis on the GUI PATH. Claude Desktop launches with a minimal environment, not your shell's. Ifnodeisn't found, use its full path (where nodeon Windows,which nodeon macOS). - Re-run
npm run authif tools report that Garmin rejected the session.
Try it
Three prompts that exercise the whole suite end to end:
Health
Pull my sleep and resting heart rate for the last five days, along with daily steps and stress. Is my resting heart rate trending up, and does it line up with the nights I slept badly?
Training
What's my current training status, VO2 max and fitness age? Compare my weekly training load against my load target range and tell me whether to push harder or back off this week.
Activity
Find my most recent run, break down the lap splits and time in each heart rate zone, and tell me whether I paced it evenly or went out too fast.
Tools
Health — get_sleep_data, get_heart_rate, get_body_battery,
get_stress_and_respiration, get_daily_summary, get_body_composition
Training — get_training_status, get_hrv_data, get_cycling_metrics,
get_training_plans_and_workouts, create_workout
Activities — list_activities, get_activity_details, update_activity,
download_activity_file
create_workout, update_activity and download_activity_file are marked as
writes, so Claude Desktop will ask before running them.
What your device actually reports
Garmin answers 200 with empty data for metrics your watch doesn't record,
rather than an error. On this account, body battery and HRV come back empty on
every date checked — the tools say so explicitly instead of returning a
confident-looking zero.
Development
npm run build # compile server and serverless handler
npm test # build, then run the assert-based checks
Tokens live at ~/.garmin-mcp/tokens.json, overridable with
GARMIN_TOKEN_PATH. See .env.example for the full list of variables.
Remote mode
The same tools can run as a hosted connector for Claude Web: a Vercel deployment with an OAuth 2.1 authorization server and Garmin tokens in Upstash Redis. Add one URL in Claude, log in, and the tools appear.
Local stdio stays the simpler option and exposes nothing to the internet. See DEPLOY.md for the hosted setup and its trade-offs.
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.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
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.