Marrow
EN
Download

Automations: your data, pushed on a schedule

Marrow can POST a file of your health data to any URL you choose, on the cadence you choose, from the phone directly. This page is the reference: how to set one up, what each payload looks like, and how to write something that receives it.

Ready-made setups for n8n, Zapier, Make, Home Assistant, Google Sheets and InfluxDB are under connections.

All Marrow guides: MCP · Self-hosted mirror · Automations · Files · Support

What an automation is

A saved job in Export, Automated: one kind of data, one format, one date range, one URL, one cadence. Marrow builds the file from its verified store and POSTs it to the URL with the app closed, after every sync commit and on each background refresh, whenever the cadence has elapsed. Nothing passes through us: the phone talks straight to your endpoint.

Setting one up

  1. Connects, Exports, Automated: turn it on, then New Automation. Made from a tool's page (n8n, Zapier, Home Assistant and the rest under connections), the automation carries that tool's name and says next to each field what the tool wants there.
  2. Address: the https URL it should POST to. Headers: any your receiver needs (an Authorization bearer token, an X-Gitlab-Token, anything). When the tool needs none, Add header is under More options.
  3. Data: Health metrics, Workouts, Strength log, Nutrition diary, or Symptoms and mood, as JSON or CSV. One automation carries one kind, so make one per kind you want. Each one that is on counts as a connection: Free keeps 3 connections on at a time, Marrow Pro up to 20, AIs and servers included.
  4. Cadence: every so many minutes, hours or days. The date range, the aggregation and the timeout are under More options.
  5. Run now sends a first payload so you can check the receiver before waiting for the schedule.

Date ranges

Payload shapes

Health metrics (daily totals, JSON)

The same structure as Health Auto Export's metrics (a metrics list of name, units and data with qty), so a receiver written for that app needs only small changes: dates here are plain days, not timestamps, and the workout and strength payloads below are Marrow's own. Values are in the stored unit named in units (kg, m, mL, degC, kcal, count). Heart rate comes as each day's Avg, Min and Max in bpm, and sleep as each night with its stages in hours. Convert on your side if you want imperial.

{"data": {"metrics": [
  {"name": "body_mass", "units": "kg",
   "data": [{"date": "2026-09-11", "qty": 81.3}]},
  {"name": "step_count", "units": "count",
   "data": [{"date": "2026-09-11", "qty": 8746}]},
  {"name": "heart_rate", "units": "bpm",
   "data": [{"date": "2026-09-11", "Avg": 71, "Min": 52, "Max": 148}]},
  {"name": "sleep_analysis", "units": "h",
   "data": [{"date": "2026-09-11", "totalSleep": 7.2, "asleep": 7.2, "deep": 1.1, "rem": 1.6, "core": 4.3, "awake": 0.2,
             "sleepStart": "2026-09-10 23:36:00 -0400", "sleepEnd": "2026-09-11 07:02:00 -0400"}]}
]}}

Raw samples (the Raw samples grouping) send every sample with its start and end time instead of a daily total. CSV metrics come as the wide table described on the files page.

Workouts

{"data": {"workouts": [
  {"activity": "Running", "start": "2026-09-11T18:43:41Z", "end": "2026-09-11T19:13:53Z",
   "duration_min": 30.2, "kcal": 512, "distance_mi": 3.06, "avg_hr": 174, "source": "Apple Watch"}
]}}

Strength log

Every working set (warm-ups are left out), plus per-exercise progress and per-muscle-group volume. Weights are in the unit you chose in Settings, Units. The unit field says which. Each set's set_type is work, failure or drop, and a set also carries rir when you logged reps in reserve. set_credits counts the sets credited to a muscle group, in fractions when an exercise works several groups.

{"data": {"unit": "lb",
  "sets": [{"time": "2026-09-11T17:02:00Z", "session": "Push day", "exercise": "Bench Press",
            "set": 1, "weight": 185, "reps": 8, "est_1rm": 229.6, "set_type": "work"}],
  "exercises": [{"exercise": "Bench Press", "best_est_1rm": 229.6,
                 "sessions": [{"date": "2026-09-11T17:02:00Z", "sets": 3, "top_weight": 185, "top_reps": 8, "est_1rm": 229.6, "volume": 3885}]}],
  "muscle_groups": [{"group": "Chest", "volume": 3108.0, "set_credits": 2.4}]
}}

Nutrition diary

{"data": {"days": [
  {"day": "2026-09-11", "entries": [
    {"meal": "breakfast", "name": "Oatmeal", "grams": 250, "kcal": 300,
     "protein_g": 10, "carbs_g": 54, "fat_g": 5, "logged_at": "2026-09-11T12:10:00Z",
     "nutrients": {"dietary_fiber": 8.0, "dietary_sugar": 1.2}}
  ]}
]}}

Writing a receiver

Anything that accepts an HTTPS POST and answers 2xx works: a webhook service, a serverless function, a script on a home server. Three rules keep it robust:

A twelve-line Python receiver that archives every post:

from http.server import BaseHTTPRequestHandler, HTTPServer
import time

class H(BaseHTTPRequestHandler):
    def do_POST(self):
        body = self.rfile.read(int(self.headers.get("Content-Length", 0)))
        with open(f"marrow-{int(time.time())}.json", "wb") as f:
            f.write(body)
        self.send_response(200)
        self.end_headers()

HTTPServer(("0.0.0.0", 8798), H).serve_forever()

Put it behind HTTPS (a Cloudflare Tunnel, Tailscale Funnel, or any reverse proxy) and point the automation at it. If what you want is a queryable copy rather than a folder of files, the self-hosted mirror is that receiver already written.

Good to know

Troubleshooting

Test with Run now. The line under it says what happened, and More options, Activity log keeps the last 30 runs with the same line. Turn on Notify when run (More options) to get a notification for each run while you check.

What Marrow showsWhat to do
A summary of what was sent, such as 12 diary entriesDelivered: the receiver answered 2xx. If nothing shows on its side, see the last rows below.
Server answered 401 or 403The receiver wants a header, or the one you added is wrong. Check its name and value, including Bearer and the space after it.
Server answered 404Wrong address, or the receiver is not listening. In n8n, use the Production URL and publish the workflow (Activate in older versions). Check for a missing path at the end.
Server answered 405The address does not take a POST. Point it at the endpoint meant for incoming data.
Server answered 413The payload is too big for the receiver. Pick a shorter date range, fewer metrics, or Daily totals in place of Raw samples.
Server answered 500 or another 5xxThe receiver failed while handling it. Look at its own log. Archive the raw body first, then parse.
The request timed outThe receiver took longer than the timeout. Answer first and work after, or raise Timeout under More options (up to 300 s).
The App Transport Security policy requires the use of a secure connectionThe address is plain http on the internet. Use https. iOS lets Marrow use plain http only on your own network.
A server with the specified hostname could not be foundA typo in the address, or a name only your home network knows. Check it in Safari on the phone.
No valid URL - set an https endpoint firstThe address is empty or does not start with https:// (or http:// on your own network).
Nothing to export in this rangeNo data of that kind in the date range. Pick a longer range under More options, or log something first.

Fixes for one tool at a time are on its page: n8n, Zapier, Make, Home Assistant, Google Sheets, InfluxDB, any URL.