ZeppBridge MCP

An MCP server over your local Zepp and Amazfit database: agents read sleep, heart rate and workouts and draft plans

  • 251 ★ on the repository
  • Last commit
  • MIT
MCP server

Broad access

Why this level

  • Reads personal health data: sleep, heart rate, stress, blood oxygen and GPS routes
  • The extended connection uses the Zepp account's email or phone and password
  • Publishing a training plan to the watch needs a separate permission and changes data on the device
All reasons and checks

lingcang728/zeppbridge

Install

Manual install

Text for your agent

First make sure the user has the ZeppBridge desktop app installed, has signed in to Zepp and has synced at least once: the CLI and MCP do not sign in. The path to zeppbridge-mcp is shown in Settings under AI tools. Connect it as a stdio server in the MCP client. If the data directory is non-standard, set ZEPPBRIDGE_DATA_DIR to an absolute path. Only draft training plans and never send them to the watch without the user's explicit permission.

Access and checks

How we review

Broad accessThe tool can change data, run commands or handle secrets.

  • Reads personal health data: sleep, heart rate, stress, blood oxygen and GPS routes
  • The extended connection uses the Zepp account's email or phone and password
  • Publishing a training plan to the watch needs a separate permission and changes data on the device

This is third-party code. Review the repository files before installing.

What it does

ZeppBridge is a desktop app for owners of Zepp and Amazfit watches. It syncs records from the Zepp cloud into a local database `data/zepp.db`, shows trends for sleep, heart rate, load and workouts, exports data and sends training plans to the watch. For agents the app ships two tools, `zeppbridge-mcp` and `zeppbridge-cli`. They are thin layers over a shared core and read the same database as the app. The MCP server gives compatible clients queries over the local records. Health data is read-only and the server does not touch the network. An agent can draft training plans, while publishing a plan to the watch needs a separate permission. The repository includes a `zepp-coach` plugin for a coaching flow over your Amazfit data. The CLI runs `status`, `sync`, `reprocess`, `export` and `contract` with no prompts or windows, and its `--json` output suits scripts. On a machine with no desktop you take the `zeppbridge-tools-<version>-<platform>.zip` archive from Releases and verify it against `SHA256SUMS.txt`. The tools resolve the data directory from their own location, so on a server you set it with `ZEPPBRIDGE_DATA_DIR`. Signing in to a Zepp account happens only in the desktop app.

Who it is for. Owners of Zepp and Amazfit watches who want to analyze sleep, heart rate and workouts with Claude or Cursor without sending a raw export to the cloud.

Good fit when

  • You want a weekly review of training and recovery: the agent reads sleep, HRV and load from the local database
  • You want a running plan built around real heart rate and pace and then sent to the watch after you confirm
  • You want to keep your Zepp history yourself: the database lives on your computer, with backups and export to JSON, CSV, GPX and FIT
  • A script or NAS scheduler should export data daily with `zeppbridge-cli export`
  • An agent on a headless server needs the data: the separate archive with CLI and MCP runs without a UI

Not a fit when

  • Your watch is another brand, such as Garmin or Apple Watch: ZeppBridge syncs only Zepp and Amazfit records
  • The Zepp phone app has not yet uploaded data from the watch: the local database fills only after a sync
  • You need third-party sign-in for extended metrics: the advanced data connection does not support third-party sign-in
  • You work on macOS or Linux and need full verification: V3 has not yet been fully verified on real machines there

Example request

Through zeppbridge-mcp look at my workouts and sleep for the last two weeks and draft a running plan for next week

MCP

Transport
stdio
Authentication
not required

Environment variables

ZEPPBRIDGE_DATA_DIR
Absolute path to the data directory when the tools do not sit next to the app

Limitations

  • Signing in to a Zepp account, including the extended data connection (HRV, stress, blood oxygen), happens only in the desktop app. The extended connection uses your Zepp email or phone and password.
  • Installers are not code-signed yet: Windows may warn about the publisher and macOS may refuse to open the app. Windows is the main development platform, and V3 on macOS and Linux has not been verified on real machines.
  • Available metrics depend on the watch, the Zepp cloud and the authorization scope. Missing data stays blank instead of being filled with zeros.
  • Upgrading from 2.x is irreversible: the old version cannot open a database after V3, so keep a copy of the database first.

How to disable. Remove the zeppbridge-mcp server from the MCP client config. The app uninstalls the usual way through the system installer, and the database stays in the data directory: delete it by hand if you no longer need it.

FAQ

How do I connect ZeppBridge to Claude or Cursor?

Install the desktop app, sign in to Zepp and wait for the first sync. Then point the MCP client at `zeppbridge-mcp`: Settings, AI tools shows its path. The README has no ready install command.

What data does the agent see?

Sleep, heart rate, HRV, steps, stress, workouts with pace and GPS routes and weekly reports, if the watch measures them and the cloud kept them. Health data is read-only and the MCP server does not use the network.

Can the agent send a plan to the watch?

It can draft a plan, and direct publishing needs a separate permission. You give the confirmation yourself, and delivery to the watch depends on the device model and sync state.

How do I run the CLI and MCP on a headless server?

Download the `zeppbridge-tools-<version>-<platform>.zip` archive from Releases, check `SHA256SUMS.txt` and point `ZEPPBRIDGE_DATA_DIR` at the database directory. Sign-in happens only in the app, so the database and token move separately.

How is ZeppBridge different from Garmin MCP?

It works with Zepp and Amazfit watches and reads a local database that it syncs itself. Garmin MCP talks to the Garmin Connect service.

Discussion

No questions or reviews yet.

Strava MCP

Strava MCP Server

Strava MCP server: an agent reads your workouts, heart rate, power, segments and routes and exports routes to GPX and TCX

MCP serverBroad access506Repository stars

Unofficial Coros MCP server: sleep, HRV, load and workouts from Training Hub, plus scheduling sessions in the calendar.

MCP serverBroad access125Repository stars
Official

An MCP server for workout tracking: Claude Desktop, Cursor and Cline read routines, sets, body weight and 1RM from your self-hosted openGym

MCP serverExtended access21.1KRepository stars
Official

Official COROS MCP: an agent reads workouts, sleep and heart rate and drafts training plans saved to your account.

MCP serverExtended access82Repository stars

A read-only MCP server that brings WHOOP recovery, sleep, HRV, strain and workouts to Claude, Codex and Cursor.

MCP serverExtended access181Repository stars

An agent reads sleep, activity and heart data from Fitbit and Pixel Watch over OAuth, with tokens kept on your computer.

MCP serverBroad access82Repository stars

Comparison with similar tools

Cards that share the main search query, and how they differ.

ToolOfficialAuthenticationRepository starsMain difference
ZeppBridge (this card)nonot required251
Garmin MCPnoAPI key1,498Garmin Connect data through an unofficial library, while ZeppBridge works with Zepp and Amazfit watches and a local database
Apple Health MCP Servernonot required575reads an Apple Health export with SQL queries, while here the Zepp database syncs itself and training plans exist

README in short

The page walks through the app by section: the metrics overview, a talk with AI through a Markdown file, turning the reply into a weekly training plan and sending it to the watch. It then covers data storage, backups, export and connecting your own tools through MCP, CLI and a local REST API. Next come installation on Windows, macOS and Linux, the first Zepp connection and where the data lives.

README badge

Report a problem