Garmin Workout Pipeline
Enables building and uploading structured workouts to Garmin Connect using YAML or natural language, supporting running, cycling, and strength with zones and circuits.
README
Garmin Workout Pipeline
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.
<p align="center"> <img src="assets/demo.gif" alt="MCP server demo — building a workout conversationally" width="800"> </p>
Why This Exists
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:
- Write workouts as code — YAML files you can version, share, and iterate on
- Build workouts conversationally — tell Claude what you want via MCP, and it handles the structure
- Push to Garmin Connect in one command — from terminal to watch in seconds
- Manage your training library — list, schedule, and delete workouts programmatically
Quickstart
# Install
pip install garmin-workout-pipeline
# Set credentials
export GARMIN_EMAIL=you@example.com
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.
Two Ways to Build Workouts
1. YAML (version-controlled, repeatable)
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 }
2. Natural Language via MCP (conversational, fast)
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.
MCP Server Setup
24 tools for full workout lifecycle management through any MCP-compatible client.
Claude Code
claude mcp add garmin-workouts \
-e GARMIN_EMAIL=your-email@example.com \
-e GARMIN_PASSWORD=your-password \
-- garmin-mcp
Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"garmin-workouts": {
"command": "garmin-mcp",
"env": {
"GARMIN_EMAIL": "your-email@example.com",
"GARMIN_PASSWORD": "your-password"
}
}
}
}
<details> <summary><strong>All 24 MCP Tools</strong></summary>
| Category | Tools |
|---|---|
| Workout | create_workout, get_workout, set_workout_name, clear_workout |
| Steps | add_warmup, add_cooldown, add_run, add_bike, add_exercise, add_rest, add_recovery, remove_step |
| Circuits | add_circuit, end_circuit |
| Garmin Connect | preview_upload, upload_workout, list_workouts, delete_workout |
| Reference | list_exercises, get_zones, validate_workout |
| Templates | save_yaml, load_template, list_templates |
</details>
CLI Reference
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
Workout Types
Running
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 }
Strength / Cardio
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 }
Cycling
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" }
Step Types Reference
| Type | End Conditions | Targets |
|---|---|---|
warmup |
duration, lap | zone, exercise |
cooldown |
duration, lap | zone, exercise |
run |
duration, distance, lap | zone, pace, hr |
bike |
duration, distance, lap | zone, power, power_pct |
recovery |
duration, distance, lap | zone |
exercise |
duration, reps, lap | — |
rest |
duration | — |
circuit |
iterations | nested steps |
Zones
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.
Installation Options
From PyPI (recommended)
pip install garmin-workout-pipeline
From GitHub
uv tool install git+https://github.com/k-schmidt/Garmin-Workout-Pipeline.git
From Source
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
Project Structure
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
Development
uv run pytest -v # run tests
uv run ruff check . --fix # lint
uv run ruff format . # format
Contributing
Contributions welcome. Open an issue or submit a PR — whether it's a new exercise, a workout template, a bug fix, or documentation improvement.
License
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。