Garminmcp

io.github.davidmosiahPublished by GitHub account davidmosiahv0.5.3Updated Oct 24, 2026

Privacy-first unofficial Garmin Connect MCP for AI health, sleep, activity & recovery agents.

The publisher's own server: Garmin MCP

VerifiedSTDIODesktop onlyLocation & LifestyleData & Analytics

Overview

AI-generated overview

Lets an assistant read your Garmin Connect health data — Body Battery, sleep, HRV, stress, activities and training readiness — from a local server.

What you can do
An unofficial local MCP server that connects an assistant to your Garmin Connect account. It exposes tools such as garmin_connection_status, garmin_daily_summary and garmin_weekly_summary, plus per-day signals (sleep, HRV, stress, Body Battery, training readiness, heart rate, SpO2, respiration, intensity minutes, hydration), activities, profile, devices and weight, along with prompts and resources. A privacy mode controls how much payload detail is returned, and an optional local SQLite cache can be enabled.
When to use it
Use it when you want an assistant to summarize your Garmin recovery, sleep and training data, build daily or weekly briefs, or investigate a specific day's stress and readiness. It suits personal wellness and training workflows, not clinical or diagnostic use.
Requirements
Runs locally over stdio (or optional loopback HTTP) via the npm package garmin-mcp-unofficial, so Node.js is needed. You must run the auth command once and log in with your Garmin email, password and MFA when prompted; tokens are then stored locally. Optional environment variables include GARMIN_TOKEN_PATH, GARMIN_PRIVACY_MODE, GARMIN_CACHE, GARMIN_CACHE_PATH, GARMIN_NO_CACHE and GARMIN_DOMAIN. Network access to Garmin Connect is required.
Before you install
Uses unofficial Garmin Connect personal-token access, not an official Garmin API partnership, and can break if Garmin changes its private auth or endpoints. Your Garmin password is prompted for during login but not stored; only tokens persist, at GARMIN_TOKEN_PATH. Repeated failed logins can trigger throttling or Cloudflare challenges, so stop retrying. It is not a medical device and gives no medical advice.

Installation

In SourceWeft

  1. Open Garminmcp in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

Upgrading to 0.8.0: read the migration guide for local-only HTTP, disabled authenticated-response caching, and stricter input/profile limits.

Garmin MCP

Give your AI agent your Garmin Body Battery, training readiness, sleep, HRV and activities.
Local-first MCP server — tokens never leave your machine.

[npm version] [GitHub release] [npm downloads] [License MIT] [Site]

[GitHub stars] [Built for MCP] [Verified release index] [Hermes one-command setup] [Garmin]

⚡ One-command install with Delx Wellness for Hermes: npx -y delx-wellness-hermes setup — preconfigures this connector and the full Delx Wellness stack in a dedicated Hermes profile.

Or wire it standalone into Claude Desktop / Cursor / ChatGPT Desktop — see the install section below. Runnable examples live in the Delx Wellness hub.

Public proof: Garmin MCP is tracked in the Delx Open Source Growth Snapshot alongside downloads, stars and next-action priorities. If it saves you Garmin Connect and MCP setup time, star this repo so other training-focused agent builders can find it faster.

First useful prompt: Use garmin_connection_status, then garmin_daily_summary, then give me a 5-line operating brief for today.


Garmin MCP

Give your AI agent your Garmin Body Battery, training readiness, sleep, HRV and activities — local-first, tokens never leave your machine.

Unofficial project. Not affiliated with, endorsed by or supported by Garmin. This is not official Garmin Health API partnership access — it uses the unofficial Garmin Connect personal-token mode.

Part of Delx Wellness, a registry of local-first wellness MCP connectors.

[Garmin MCP local-first agent workflow demo]

Quickstart in 60 seconds

No Garmin developer app is required. setup only writes local MCP configuration; it does not ask for your Garmin password.

bash
npx -y garmin-mcp-unofficial setup            # writes local confignpx -y garmin-mcp-unofficial auth             # built-in login, prompts for credentials locally (no Python needed)npx -y garmin-mcp-unofficial doctor           # verifies you're ready

HTTP (v2 stateless)

Default is stdio. Optional Streamable HTTP — no session id, JSON responses, loopback only:

bash
npx -y garmin-mcp-unofficial --http# GET  http://127.0.0.1:3000/health# POST http://127.0.0.1:3000/mcp   (sessionless)

Env: GARMIN_MCP_HOST, GARMIN_MCP_PORT, GARMIN_MCP_TRANSPORT=http.

Or one shot: npx -y garmin-mcp-unofficial setup --auth

auth runs a self-contained Node login and prompts locally for Garmin email, password and MFA when needed. The MCP does not store your Garmin password — only Garmin Connect tokens, saved at ~/.garmin-mcp/garmin_tokens.json with user-only permissions. See the auth quickstart walkthrough for real terminal output, or docs/quickstart.md for the full flow.

If Garmin returns HTTP 429, a Cloudflare challenge, or an auth message that says Garmin SSO omitted responseStatus.type, stop retrying for a while. Repeated headless login attempts can make the private endpoint throttle harder. Use docs/auth.md for the safe recovery path.

Then add this to your MCP client config:

json
{  "mcpServers": {    "garmin": {      "command": "npx",      "args": ["-y", "garmin-mcp-unofficial"]    }  }}

Try it with your agent

text
Use garmin_connection_status to check setup, then run garmin_daily_summary.Give me a 5-line operating brief for today.
text
Call garmin_weekly_summary with response_format=json. Identify my biggestrecovery/sleep/stress bottleneck and give me a next-week plan.
text
Use the garmin_intraday_investigation prompt for date=today, focus=stress.Don't claim Garmin can prove anything it can't.

Tools

Start with garmin_connection_status, then garmin_daily_summary (daily readiness, sleep, load) or garmin_weekly_summary (scorecard, bottlenecks, next-week plan). The server also exposes per-day signals (sleep, HRV, stress, Body Battery, training readiness, heart rate, SpO2, respiration, intensity minutes, hydration), activities, profile/devices and weight, plus prompts and resources.

See docs/tools.md for the full tool list, prompts, resources, data-availability matrix, configuration, Hermes setup and development notes.

Privacy & what runs offline

  • GARMIN_PRIVACY_MODE defaults to summary (more conservative than other Delx Wellness connectors) because the auth model is unofficial.
  • In structured mode, normalized aliases are additive: complete upstream Garmin fields and nested DTOs remain available after secret/GPS redaction.
  • Activity date ranges preserve the supplied calendar date and reject invalid values before contacting Garmin Connect.
  • Garmin Connect tokens are stored at ~/.garmin-mcp/garmin_tokens.json with user-only permissions and are never returned by tools. Your Garmin password is never stored — only short-lived tokens persist locally.
  • The MCP client never sees Garmin credentials or tokens. Authenticated Garmin responses are never cached; GARMIN_CACHE has no effect.
  • This is not medical advice. The server exposes user-authorized data for personal AI workflows, not diagnosis or treatment.

See docs/privacy.md for the full privacy model.

See the full agent demo →

Want to see a connector like this drive a real end-to-end decision? The shared, reproducible proof lives in delx-living-body:

bash
npx -y delx-living-body demo

It answers the anchor question — "Should I train hard today?" — by combining recovery, sleep and training-load signals across the Delx Wellness connectors.

See also

The full Delx Wellness connector library:

One-command setup for Hermes — preconfigures every connector above plus wellness skills + onboarding: delx-wellness-hermes.

Related local-first health MCP

Peer (not a Delx package): mi-fitness-data-bridge by Kindred / @shkyyy18 — Xiaomi Mi Fitness, local-first, shares the agent-safe-series/v1 dense series envelope (duration-anchored coverage, hard max_points, full-res stats, no GPS in series tools). Design log: #19 · parity notes: docs/agent-safe-series-kindred.md.

📧 Contact & Support

License

MIT — see LICENSE. Code of Conduct.

Disclaimer

This software is provided as-is. It is not a medical device, does not provide medical advice, and should not be used for diagnosis or treatment. The unofficial Garmin Connect mode can break if Garmin changes private auth or endpoints. Always consult qualified professionals for medical concerns.

Daily brief demo: docs/daily-brief-demo.md

Skill or MCP

Same package, two doors. MCP registers tools on stdio/HTTP. The skill can drive the same tools through the CLI when the client has no MCP:

bash
npx -y garmin-mcp-unofficial call garmin_connection_status --json '{}'

Copy skill/SKILL.md into your agent skills dir.

Source: README.md at commit 123c5a1

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v0.5.3LatestSep 16, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts

Other Garminmcp integrations