接入你的智能体
从发布者处安装。不是 PlugMCP 端点。
远程 URL 还是本机命令? 两种安装方式的区别
Stop clicking through Garmin Connect's UI to build every workout. Define workouts in YAML, or just tell Claude what you want in plain English — then push them straight to your watch.
Garmin Workout Pipeline is an open-source CLI and MCP server that compiles structured workout definitions into Garmin Connect API payloads. It supports running, cycling, and strength/cardio workouts with pace/HR/power zones, 84+ exercises, circuits, and weekly scheduling.
"Build me a Hyrox sim with 8 stations, 1km runs between each, and a 10-minute warmup."
That's a real prompt. The MCP server turns it into a fully structured Garmin workout and uploads it to your watch.
If you've ever built a complex interval workout in Garmin Connect, you know the pain: endless dropdowns, no copy-paste, no version control, and good luck reusing that Hyrox sim you spent 15 minutes clicking together.
This tool lets you:
# Install
pip install garmin-workout-pipeline
# Set credentials
export [email protected]
export GARMIN_PASSWORD=your-password
# Push a workout to your watch
gwp push workouts/templates/hyrox-sim.yaml --zones workouts/zones.yaml
That's it. Workout is on Garmin Connect, ready to sync to your device.
name: "Threshold Intervals"
type: running
steps:
- warmup: { duration: "10:00", zone: easy }
- run: { distance: "1km", pace: { min: "6:25/mi", max: "6:40/mi" } }
- recovery: { duration: "2:00" }
- run: { duration: "5:00", zone: threshold }
- cooldown: { duration: lap, zone: easy }
Connect the MCP server to Claude Desktop or Claude Code, then just describe what you want:
"Create a 5x1km workout at threshold pace with 2-minute recoveries, 10-minute warmup and cooldown"
Claude builds the structured workout, previews it, and uploads it — all through conversation.
24 tools for full workout lifecycle management through any MCP-compatible client.
claude mcp add garmin-workouts \
-e [email protected] \
-e GARMIN_PASSWORD=your-password \
-- garmin-mcp
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"garmin-workouts": {
"command": "garmin-mcp",
"env": {
"GARMIN_EMAIL": "[email protected]",
"GARMIN_PASSWORD": "your-password"
}
}
}
}
gwp push <file> --zones <zones.yaml> # Upload workout to Garmin Connect
gwp push <file> --zones <zones.yaml> --schedule 2026-04-29 # Upload + schedule
gwp push <file> --zones <zones.yaml> --dry-run # Preview JSON without uploading
gwp validate <file> --zones <zones.yaml> # Compile and validate only
gwp list # List workouts on Garmin Connect
gwp delete <workout-id> # Delete a workout
gwp zones --zones <zones.yaml> # Show resolved zone values
Pace targets, HR zones, distance and time-based intervals.
name: "Speed 400s"
type: running
steps:
- warmup: { duration: "10:00", zone: easy }
- run: { distance: "400m", pace: { min: "5:30/mi", max: "5:45/mi" } }
- recovery: { duration: "1:30" }
- cooldown: { duration: "10:00", zone: easy }
84 exercises with rep counts, weights, and circuit support.
name: "Hyrox Strength"
type: strength
steps:
- warmup: { duration: lap, exercise: rowing_machine }
- circuit:
iterations: 4
steps:
- exercise: { exercise: wall_ball, reps: 20, weight: 13 }
- exercise: { exercise: weighted_lunge, reps: 20, weight: 45 }
- rest: { duration: "2:00" }
- cooldown: { duration: lap, exercise: rowing_machine }
Power zones, FTP percentages, and duration-based blocks.
name: "Sweet Spot"
type: cycling
steps:
- warmup: { duration: "10:00", zone: z2 }
- bike: { duration: "20:00", zone: threshold }
- cooldown: { duration: "5:00" }
Define your training zones once in workouts/zones.yaml with HR, pace, and power targets per sport. The compiler resolves zone names like threshold, z2, and easy to Garmin API target values.
pip install garmin-workout-pipeline
uv tool install git+https://github.com/k-schmidt/Garmin-Workout-Pipeline.git
curl -LsSf https://astral.sh/uv/install.sh | sh
git clone https://github.com/k-schmidt/Garmin-Workout-Pipeline.git
cd Garmin-Workout-Pipeline
uv sync
garmin_pipeline/
mcp_server.py # MCP server for Claude Desktop/Code
cli.py # Click CLI (gwp command)
compiler.py # Workout model → Garmin API JSON
exercises.py # Exercise name → Garmin category/name registry
loader.py # YAML parser with !include support
models.py # Pydantic workout models
sync.py # Garmin Connect auth and upload
zones.py # Zone resolution (HR, pace, power)
workouts/
zones.yaml # Training zone definitions
templates/ # Workout YAML files
tests/
fixtures/ # Golden reference JSON
test_compiler.py # Compiler golden tests
test_loader.py # YAML loading and !include
test_models.py # Step parsing
test_zones.py # Zone resolution
uv run pytest -v # run tests
uv run ruff check . --fix # lint
uv run ruff format . # format
Contributions welcome. Open an issue or submit a PR — whether it's a new exercise, a workout template, a bug fix, or documentation improvement.
Build and upload Garmin Connect workouts from YAML or natural language. CLI + MCP server with pace/HR/power zones, 84 exercises, circuits, and scheduling. Works with Claude Desktop and Claude Code.
工具
暂无工具列表。远程服务器在首次健康检查后会获得工具列表。
常见问题
在本页复制适用于 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 配置。它使用发布者的安装方式,而不是 PlugMCP 代理。
此记录未列出必需的环境变量。
Garmin Workout Pipeline 通过 stdio 在本地运行。
上次检查时间:2026-10-28。 Build and upload Garmin Connect workouts from YAML or natural language. CLI + MCP server with pace/HR/power zones, 84 exercises, circuits, and scheduling. Works with Claude Desktop and Claude Code.