Skip to main content
Glama
MatheuslFavaretto

amazfit-mcp

amazfit-mcp

CI

MCP server that turns Claude into your personal training analyst — crossing your training spreadsheet (strength) and Amazfit watch (running + recovery via Apple Health). No frontend, no hosting: just data exposed as tools; Claude does the analysis.

16 tools: weeks, sessions and exercise progression from the spreadsheet · runs with pace/HR/TRIMP · resting HR, HRV and sleep · ACWR and daily CTL/ATL/TSB (form) · a transparent 0–100 readiness score · daily notes exported to Obsidian · a self-contained HTML health report.

Setup

python3.12 -m venv .venv && .venv/bin/pip install -e .

claude mcp add amazfit-mcp \
  --env PYTHONPATH="$PWD/src" \
  --env AMAZFIT_MCP_XLSX="$PWD/data/planilha.xlsx" \
  --env AMAZFIT_MCP_RECOVERY_DIR="$HOME/Library/Mobile Documents/com~apple~CloudDocs/HealthAutoExport" \
  -- "$PWD/.venv/bin/python" -m amazfit_mcp

Optional envs: AMAZFIT_MCP_OBSIDIAN_DIR (vault folder for the daily notes), AMAZFIT_MCP_HR_MAX / AMAZFIT_MCP_HR_REST (your real HRs, to calibrate TRIMP).

Watch data (one-time, on the iPhone): Zepp app → Profile → Apple Health → enable HR, resting HR, Sleep, HRV. Then install Health Auto Export, create a daily JSON automation (those metrics + Workouts) into an iCloud folder — the MCP reads it from there. Validate by asking Claude for recovery_status.

Already have a CSV dump? If you exported everything from Health Auto Export as CSV instead of the JSON automation, backfill the store with:

python -m amazfit_mcp.hae_csv ~/Downloads/HealthAutoExport_<date>

Related MCP server: Strava MCP Server

Use

Just ask Claude:

"how was my Saturday of week 1?" · "am I progressing on the squat?" · "how is my form?" · "am I ready to train hard today?" · "export this week's notes to Obsidian"

Development

.venv/bin/pip install -e .[dev]
.venv/bin/python -m pytest        # 94 tests
.venv/bin/ruff check . && .venv/bin/mypy   # lint + types (same gate as CI)

Architecture, spreadsheet layout and design decisions: CONTEXT.md. There is also an experimental Zepp cloud extractor: python -m amazfit_mcp.zepp_cloud --help.

Contributing

Open source, MIT. Issues and PRs welcome — keep tests green and add one for what you change. Good first ideas: extractors for other watches that write the Health Auto Export format into the store (zero reader changes), stress/PAI in the Zepp extractor, more sports in workouts.py. If this helped you, a ⭐ helps others find it.

Related MCP Connectors

Related MCP Servers

  • F
    license
    Not graded
    quality
    B
    maintenance
    Connects Claude to Garmin Connect data for personalized running coaching, including morning readiness checks, post-run analysis, weekly reviews, and goal tracking.
    -
  • A
    license
    Not graded
    quality
    D
    maintenance
    Exposes Whoop fitness data (recovery, sleep, strain, workouts) to Claude for use as a daily training coach, enabling natural language queries about your health metrics and training readiness.
    MIT