Privacy

What Suppermode stores

The content is yours: the recipes you save, every version of them, and the notes you write about cooking them.

And your cook profile, which is worth naming on its own rather than folding into "content": what equipment you have, who you are feeding, what you cannot eat, what you like, and how much time and confidence you have. You write it, none of it is required, and the line about what you cannot eat is the most sensitive thing this service holds. It exists so an assistant can adapt a recipe without you explaining your kitchen again every conversation.

Alongside that, the ordinary things an account needs: your email address, when you joined, and your password stored as a one-way hash that nobody can read back. The address is what you sign in with and where a password reset goes, and it is not used for anything else. Product analytics are described below.

And your network address, which is read on every attempt to sign in or to use a tool. What is kept against it is narrower than that, and the difference is worth stating: a counter is retained for sign-ins that failed, for requests whose credentials were rejected, and for clients registering themselves — the things worth making expensive rather than free. A tool call that succeeds is counted against your account instead. The counters sit in Suppermode's own database, they are never attached to the product analytics, and nothing here builds a picture of you out of them.

Your JSON export contains your email address, your join date, and everything you have written — every recipe, every version, every cook, and your cook profile — along with the link for any recipe you are sharing. It does not contain the password hash, which is not yours to need, or the analytics, which are held by the provider described below.

What it does not store

Suppermode does not receive your ChatGPT conversations. What reaches it is what your assistant sends when it uses one of Suppermode's tools: the recipe or note you are saving, the words you searched with when you asked it to find something, and the profile answers you confirm — along with a locale, a coarse location, and anonymised identifiers. Never your chat history, and never the conversation around those calls.

Product analytics record that an event happened and which recipe it concerned, and when you answer a profile question, which question it was. They never carry the text of your notes, your recipes, or your profile answers. They exist to answer two questions. Whether people get as far as connecting Suppermode at all, which is why a signup or a connection abandoned halfway is recorded. And whether anybody comes back to log a second cook and change a recipe because of it, which is the assumption the whole product rests on and which we would otherwise be guessing at.

When you share a recipe, that is recorded too, and so is each time somebody opens the link — counted against your account, as the recipe's owner. Nothing is recorded about the person who opened it: no account, no cookie, no identifier.

What it sends back

The other direction, which matters as much and is easier to leave unsaid. When your assistant asks Suppermode for something it gets back what it asked for: a recipe and its versions, the cooks logged against them with your observations and your ratings, the reason recorded for each change, and the cooks a change was based on. When it asks for your profile — which it is told to do before writing or adapting a recipe — it gets your answers, including the dietary line.

Alongside those, the identifiers it needs in order to refer to one of them in a later request: which recipe, which version, which cook.

It never receives your email address, your join date, how your account is secured, or any internal bookkeeping. Those are not withheld as a courtesy; they are simply not among the things the tools return.

The assistant you connect it to

That boundary runs one way, and the other way is not ours to control. When ChatGPT reads a recipe, a cooking note or your cook profile out of Suppermode, what it then does with what it read is governed by OpenAI's terms and not by this page. As their help centre described it in October 2026, a Free, Plus, Go or Pro account may have information an app returned used to train their models when the account's own "Improve the model for everyone" setting is on, while Business, Enterprise and Edu accounts do not by default. That is their policy rather than ours, so read it there for what is true today; the part that will not change is that what ChatGPT does with what it read is between you and them.

That setting lives in ChatGPT rather than here, and we cannot change it on your behalf. It is worth knowing because the cook profile is the part of this that holds dietary constraints, and because disconnecting Suppermode in ChatGPT stops any further reading while leaving your cookbook here untouched.

Your data is yours

You can export everything as JSON from your account page. The export is complete: your email address and join date, every recipe, every version, every cook you have logged, where each recipe came from, the link for any recipe you are sharing, and your cook profile.

You can delete your account from the same place. Deleting it removes your account, your cook profile, and every recipe, version and cooking note from Suppermode's live database at once. One qualification, because the honest version is better than the clean one: the database is backed up daily and those backups are kept for seven days, so a deleted cookbook is gone from the running service immediately and out of the backups within a week.

There is a second place your words can end up, and it is fair to say so plainly. Suppermode uses an error tracker, so that when something breaks we find out rather than waiting for you to tell us. A crash report includes what the code was working on at the moment it failed — which can mean the text of a recipe you were saving, or part of your cooking profile. It is not read unless something has gone wrong, it is never used to build a picture of you, and it is the reason a bug in your cookbook can be fixed the same day instead of guessed at.

To be precise about one thing rather than overclaim: the product analytics described above are held by our analytics provider, and deleting your account does not reach into their records. What they hold is a user identifier and recipe identifiers against event names — never your recipe text or your notes — so those records become identifiers pointing at data that no longer exists. The same is true of any crash report: deleting your account does not reach into the error tracker either.

If you would like either erased as well, email [email protected] and we will do it.

Sharing a recipe

Your recipes are private unless you share one. Sharing gives that recipe a link, and anyone holding the link can read the recipe as it is now — its ingredients, method and notes, updated whenever you change it. They cannot see your cooks, your cook profile, your other recipes, or who you are. The page is kept out of search engines; it is for whoever you send it to.

Stopping sharing breaks the link at once, and archiving the recipe does the same. Sharing it again makes a new link, so the old one never works again.

How long it is kept

Your account and your cookbook do not expire. Your email address, your join date, your recipes, every version of them, your cooks and your cook profile all stay until you delete them or delete your account. Nothing ages out on its own and nothing is removed for being old.

Everything else has an outer bound, and they are different enough to be worth stating one at a time. Database backups: seven days. Crash reports: ninety days, after which the error tracker discards them whether or not anyone has looked. Product analytics are the long one — our analytics provider keeps them to its own schedule, which runs to seven years.

The rate-limiting counters are the awkward one, and rounding it off would be the easy thing to do. A counter stops counting within the hour. The row holding it is cleared when that address is next seen or when the table is next trimmed, rather than at a fixed time — so an address seen once can sit in that table longer than the hour its counter was good for.

As the section above says, deleting your account does not reach the analytics or the crash reports. If you would rather not wait out either period, email [email protected] and we will erase them.

Who holds your data

Suppermode is operated by Pleasant Street Partners, LLC, a Delaware limited liability company, and that company is responsible for the data described above. Four providers each hold some part of it, for one job and nothing else: DigitalOcean runs the servers and the database, so everything above sits at rest there; an email provider delivers sign-in and password-reset mail, and the short notes that tell us when you join, connect an assistant, or log your first cook, so it handles your email address and what those messages say, including the one-time links in them; PostHog holds the product analytics; and Sentry holds the crash reports.

ChatGPT is not one of them. It is a service you connect yourself, so what it reads is covered by the section above rather than by any arrangement we have made on your behalf.

Getting in touch

Email [email protected].