SunStrong / mySunPower cloud data downloader (not local!)

This script lets you download your full solar history (whatever is left, anyway) from SunStrong cloud servers straight to a CSV file -- no emulator, no rooting, no browser. You log in with the same username/password you use in the app, and the tool pulls data from the same private API the app uses. This will NOT connect to your local PVS5/PVS6, and is firmware agnostic. It is purely to download your historical data from SunStrong's servers. It was last tested on 7/16/2026.

What you can download

Command	Scope	Granularity	Identity	History*
power	Whole system	hourly (5/15 mins don't work)	site	back to system start
energy	Whole system	day / month / year totals	site	back to system start
panels	Per panel	hourly power and energy	microinverter serial	~2 years

panels also writes per-panel daily totals and the physical layout (x/y/rotation/azimuth).

*History depth varies by account. SunPower had some database crashes a few years ago, so some of your data may have been wiped out. Also, panel-level seems to go back only a few years, so YMMV.

When SunStrong acquired these assests, they reduced system granularity from 5/15 minute intervals to 1 hour (at least for my system).

Setup

pip install requests

Credentials via environment (.env) (or you'll be prompted):

export SUNSTRONG_USERNAME="you@example.com"
export SUNSTRONG_PASSWORD="your-app-password"     # Windows PowerShell: $env:SUNSTRONG_PASSWORD="..."

Use

1. Probe first — logs in, lists your site(s), and confirms the panel date format:

python sunstrong_downloader.py probe

Note your siteKey and account timezone from the output. It writes raw_sample.json so you can eyeball the real response shape.

2. Download. Examples (adjust dates; your system's install date is a good --start):

# System hourly power/production, full history
python sunstrong_downloader.py power  --start 2019-01-01 --interval hour --out power_hourly.csv

# System energy totals per month (or day/year)
python sunstrong_downloader.py energy --start 2019-01-01 --interval month --out energy_monthly.csv

# Panel-level, hourly power+energy per microinverter serial (last ~2 years)
python sunstrong_downloader.py panels --start 2024-01-01 --out panels_hourly.csv
#   -> panels_hourly.csv, panels_hourly_daily.csv, panels_hourly_site_hourly.csv, panels_hourly_layout.csv

Dates are treated as site-local. Pick a specific site with --site <siteKey> if you have more than one. If a power/energy call returns bad query params, your interval must be lowercase.

Output columns

• power — timestamp, production_kw, consumption_kw, storage_kw, grid_kw (at --interval hour, each production_kw value equals that hour's kWh, so this doubles as system hourly energy — validated to match the energy monthly totals within ~1%)
• energy — bucket_start, production_kwh, consumption_kwh, net_grid_import_export_kwh, net_storage_kwh
• panels (hourly) — timestamp, serialNumber, power_w, energy_kwh
• panels (daily) — date, serialNumber, dailyEnergyProduction_kwh, sevenDayAverage, peakTimestamp, peakValue, energyColorCode, lastCommunicationTimestamp
• panels (layout) — serialNumber, xCoordinate, yCoordinate, rotation, azimuth, orientation

What needs a paid subscription

• Without a subscription: the panels command still works — per-panel and system production at hourly resolution (last ~2 years). power and energy return empty/zero.
• With an active subscription: power (system hourly, full history) and energy (totals, full history) light up — reaching back to system commissioning, far deeper than panel data.

So a subscription is mainly worth it to get system history older than the panel-level window. Note the tool needs the interval lowercase on those two commands (the paid API validates it).

Notes & caveats

• Calibration: dates are sent as UTC. If the very first/last buckets look shifted, your data may be keyed to site-local time — check probe's raw_sample.json and adjust --start/--end by your offset. The panels --date-format can be switched to the ISO fallback if probe says so.
• Be gentle: downloads run one chunk/day at a time with retries and backoff. Pulling many years of 5-minute data is a lot of requests — expect it to take a while. storage/grid are blank if you have no battery/consumption metering.
• Credentials are sent only to edp-api.edp.sunstrongmonitoring.com/v1/auth/okta/signin; the access token stays in memory and is never printed or written to disk.
• Blending in: the tool does not advertise itself. curl_cffi impersonates Chrome, so both the TLS handshake and the User-Agent look like an ordinary Chrome browser (a UA that can't be blocked wholesale). It sends no custom/identifying headers. (This is also required for the GraphQL host, which drops Python's default TLS fingerprint outright.)
• Endpoints are undocumented and may change; if a call starts failing, the app may have been updated.