Open Health Sync — Claude Code MCP server for wearable health data
The Open Health Sync MCP server connects to Claude Code and other clients for read-only access to wearable data. Agents can request heart rate, sleep, workouts, nutrition and daily summaries; the server runs on your own host.
AGPL-3.0 ★ 72 updated 2026-10-23
claude mcp add --env OHS_MCP_URL=<OHS_MCP_URL> --env OHS_MCP_KEY=<OHS_MCP_KEY> --transport stdio open-health-sync -- docker run -i --rm -e OHS_MCP_URL -e OHS_MCP_KEY ghcr.io/soma-intelligence/open-health-sync:0.2.0What Open Health Sync does
Open Health Sync pulls data from eighteen wearable and health sources into one model, then exposes it through a read-only MCP server. An agent can ask for heart rate, sleep, steps, workouts and nutrition without write access. Sources include Apple Health, Withings, Oura, WHOOP, Fitbit, Polar, Strava, Garmin, Dexcom, OMRON and Health Connect.
The project is a single Go binary plus Postgres. It runs on a Raspberry Pi. Besides MCP, the outputs include a REST API, FHIR R4, Open mHealth and CSV/Parquet exports. Version 0.2.0 is an early preview: APIs, config and schema can change in any 0.x release.
It ships as a Docker image ghcr.io/soma-intelligence/open-health-sync:0.2.0. Connecting an MCP client requires OHS_MCP_URL and OHS_MCP_KEY environment variables pointing at a running instance. A hosted beta with the same API exists, so you can skip running a server and get a key in a minute.
What it helps with
- give an AI agent read-only access to wearable data
- merge heart rate, sleep and steps from multiple sources into one model
- export health data to CSV or Parquet for analysis
- run a health API on your own server without third-party services
- get FHIR R4 and Open mHealth output from one binary
Example prompts for the agent
- Show me my average heart rate over the last seven days
- Compare sleep from Oura and WHOOP for the past week
- Export steps for this month to CSV and plot them
How to install Open Health Sync
Platform CLI
claude mcp add --env OHS_MCP_URL=<OHS_MCP_URL> --env OHS_MCP_KEY=<OHS_MCP_KEY> --transport stdio open-health-sync -- docker run -i --rm -e OHS_MCP_URL -e OHS_MCP_KEY ghcr.io/soma-intelligence/open-health-sync:0.2.0Replace <…> with your own values
Platform CLI
codex mcp add open-health-sync --env OHS_MCP_URL=<OHS_MCP_URL> --env OHS_MCP_KEY=<OHS_MCP_KEY> -- docker run -i --rm -e OHS_MCP_URL -e OHS_MCP_KEY ghcr.io/soma-intelligence/open-health-sync:0.2.0Replace <…> with your own values
Config file
{
"mcpServers": {
"open-health-sync": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"OHS_MCP_URL",
"-e",
"OHS_MCP_KEY",
"ghcr.io/soma-intelligence/open-health-sync:0.2.0"
],
"env": {
"OHS_MCP_URL": "<OHS_MCP_URL>",
"OHS_MCP_KEY": "<OHS_MCP_KEY>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"open-health-sync": {
"type": "stdio",
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"OHS_MCP_URL",
"-e",
"OHS_MCP_KEY",
"ghcr.io/soma-intelligence/open-health-sync:0.2.0"
],
"env": {
"OHS_MCP_URL": "<OHS_MCP_URL>",
"OHS_MCP_KEY": "<OHS_MCP_KEY>"
}
}
}
}Replace <…> with your own values
Platform CLI
gemini mcp add -e OHS_MCP_URL=<OHS_MCP_URL> -e OHS_MCP_KEY=<OHS_MCP_KEY> open-health-sync docker run -i --rm -e OHS_MCP_URL -e OHS_MCP_KEY ghcr.io/soma-intelligence/open-health-sync:0.2.0Replace <…> with your own values
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Open Health Sync
How do I install Open Health Sync in Claude Code?
Run `claude mcp add --env OHS_MCP_URL=<OHS_MCP_URL> --env OHS_MCP_KEY=<OHS_MCP_KEY> --transport stdio open-health-sync -- docker run -i --rm -e OHS_MCP_URL -e OHS_MCP_KEY ghcr.io/soma-intelligence/open-health-sync:0.2.0`. Docker needs to be available, and the environment variables must point to a running Open Health Sync instance.
What does Open Health Sync do?
It pulls data from eighteen wearable and health sources into a single model and serves it through a read-only MCP server, REST API, FHIR R4 and Open mHealth. For an AI agent it is a way to read heart rate, sleep, steps and workouts without write permissions.
What do I need to run it?
A Go binary and Postgres, or the Docker image. It runs on your own hardware, including a Raspberry Pi. The MCP client needs OHS_MCP_URL and OHS_MCP_KEY.