Effective October 15, 2026
What changed: Anthropic and GitHub are now named as sub-processors, with what each receives; the app now sends short error reports when a screen fails (section 1, Error reports); and section 1 now says accurately how long server logs are kept. Naming the providers describes a practice that already existed — scan images have gone to Anthropic and in-app reports to GitHub since those features shipped — while error reports are new. Both are material changes to this policy, so it is published under a new effective date rather than edited in place. The version in force until then is at /privacy/2026-08-29/.
Tidalyst holds a farm's operational record, and that record is commercially sensitive — a farm's yields, costs, and buyers are exactly what a competitor would most like to see. This policy says what we collect, why, who can reach it, and what we will never do with it. It is written by the people who built the controls it describes.
If a screen in the Tidalyst app fails, the app sends us a short report so we can find and fix the fault without waiting for you to tell us. A report contains the error message (shortened to a few hundred characters), a technical error code, the address of the page where it happened (never search terms or anything typed on the page), the version of the app you were using, and the internal identifiers of your farm account and your user account. It never contains anything you entered: no notes, measurements or form contents. If your phone has no signal, a few recent reports wait on the device for up to three days and are sent when signal returns; older ones are discarded. Reports are stored in Tidalyst's own database, can be read only by Tidalyst's founders (never by other farms), are used only to diagnose and fix faults, and are deleted automatically 30 days after we receive them.
What we deliberately do not collect: payment card or bank details (they go to Stripe directly and never touch our servers), and behavioral tracking of any kind. The application ships no analytics, no session recording, no advertising SDKs, and we do not sell your data or share it for marketing. Two features do send farm data to a third-party AI provider as part of doing their job — reading a scanned grade report, and helping us triage a bug report you send us. Both are named, with exactly what each receives, in the table in section 5.
To provide the service, to respond when you ask for help, and to send the transactional email the service requires (invitations, password resets, receipts). We do not send marketing email to farm members, and we do not use your data for advertising.
Every farm is a separate organization, and isolation is enforced by the database itself through row-level security: every query runs as the actual signed-in person, and the database returns only rows belonging to farms they are a member of. This holds even if application code has a bug, and it is covered by automated tests that run on every change. One farm cannot see another's data, in any form attributable to it.
Honestly: an administrator holding the database's secret key can read any row — that is true of every hosted database product. Today that access is held by the founders, by the automated systems that deploy and operate the service, and — this early in the company's life — by software agents the founders direct to fix bugs or correct records. None of those reads are logged yet; logging every one of them, attributable to a person and a reason, is a commitment we are actively building toward and will state here as fact once it ships, not before. We do not read a farm's data except to act on that farm's own support request, or to correct an error in its records. Our commitments — never selling data, never sharing it for marketing, never showing it to another farm — are contractual in our Terms of Service, section 4.
| Provider | Role | What it reaches |
|---|---|---|
| Supabase | Database and authentication | All operational data, encrypted at rest |
| Vercel | Application hosting | Data in transit while rendering pages |
| Stripe | Payments | Billing contact and payment details; no farm operational data |
| Resend | Transactional email delivery | Recipient addresses and email content |
| Cloudflare | Domain and email routing | Mail addressed to @tidalyst.com |
| Anthropic | Two AI features | When an owner or manager taps Scan to read a dealer's grade report, the image or PDF goes to Anthropic's API to pull out quantities and prices. Separately, when anyone submits in-app feedback, the report text and the technical details below are read by an automated assistant, running on Anthropic's API, that helps our team triage it. |
| GitHub | Bug-report tracking, source control, and encrypted backup storage | Every in-app feedback report becomes an issue in a private GitHub repository, carrying the report text, the page it was sent from, the reporter's first name and a short account handle (not their email address), their role on the farm, and a short list of recent technical error messages from their browser (never page content, form values, or a farm's operational numbers). Kept as long as the issue exists — we don't yet have an automatic deletion window for these, tracked as a commitment to close. GitHub's automation also holds a database credential to run scheduled jobs, and stores a nightly backup of the database and file storage — but that backup is encrypted before it leaves our server, so GitHub itself cannot read it. |
| NOAA (NDBC, CO-OPS) | Public water and tide data, polled in | Nothing — the connection is one-directional and no farm data is sent |
Adding a provider that can reach farm data is a change to this policy, and farms will be told before it takes effect. Anthropic and GitHub's access described above predates this update to the policy — we are correcting the record rather than announcing something new — and both rows have been true since the features they describe shipped.
Data is encrypted in transit (TLS) and at rest by our database provider. Said plainly, because it is routinely oversold: encryption at rest protects against a stolen disk, not against someone holding valid credentials — which is why the isolation in section 3 and the access limits in section 4 are the controls that matter. Passwords are never generated for you or sent by email; sign-in links are single-use and expire.
We keep your data as long as your organization has an account. An owner or manager can download a complete copy of the farm's operational records, as CSV files, from inside the application at any time — no request needed. On written request we will, within 30 days, also provide anything self-service export does not cover (account and billing administration, for instance), or delete your organization's data, subject to records we are legally required to keep. Closing a subscription does not delete records; deletion happens only when you ask for it.
The service is for businesses and is not directed to children under 13; we do not knowingly collect their information.
If this policy changes materially, the owner account of every farm gets email notice before the change takes effect. Questions and requests: legal@tidalyst.com.
Tidalyst LLC · Chiefland, Florida