Privacy policy

Who is responsible

Jan Vančák (IČO 87592193) is the data controller for the personal data described here, and can be reached at czechspeed1337@gmail.com.

Registered address: Na Pančavě 530, 538 51 Chrast, Česká republika.

What is stored

• Your account: the email address or social login you signed up with. This is held by Clerk, our authentication provider; the app itself never stores your password.
• Your profile: the preferred name, pronouns, age bracket, focus areas, communication style, prior-therapy answer, language and time zone you gave during onboarding.
• Your reflections: the full conversation of each reflection, its summary, the themes drawn from it, the moods you recorded before and after, and whether you found it helpful.
• Your mood check-ins, including any note you attach to one.
• Your journal notes: the title and text of each note, whether you marked it important, and whether you let Maxi use it.
• Your membership: which plan you have, whether it is a free trial, and when it started, renews or ends. Payment details never reach Maxi — Apple or Google takes the payment.
• Your AI replies: how many AI replies you have received from your free replies or from your plan in each billing period, and, for every request Maxi sends to an AI provider, how many tokens it used, its estimated cost and how long it took. Never the text of the request or of the reply.
• Your Future reports, if you write any: the patterns computed from your own records, the wording of each section, which AI provider worded it, and when it was written. Stored encrypted like your reflections, included in your export, and erased at once when you delete one.
• Safety events: when a message triggers the safety check, we record the time, a risk score, and a one-way fingerprint of the message. The message itself is not stored in that record.
• Reports you send about one of Maxi’s replies: the reason you chose, the note you added if any, and which reply it was. The reply itself is not copied into the report.
• Diagnostics: if the app or the server hits an error, a crash report goes to Sentry. Crash reports carry technical details, not the content of your reflections.

What it is used for

Your reflections and profile are used to hold the conversation, to write the summary you see afterwards, to show your trends, and to run the safety check that offers crisis resources. That is the whole list.

Your membership record is used for one thing: deciding which parts of Maxi you can use.

Your reply count decides how many AI replies you have left. The token and cost records are used only to keep the cost of running Maxi in check, and to notice when something is using far more than a person could.

The one addition is a reply you report yourself: the people who run Maxi then review that reply — and, where it is needed to understand it, the messages just before it — to find out what went wrong. A report is never sent to the AI provider.

There is no advertising in this app, no analytics profile is built about you, and nothing is sold or handed to a data broker.

The AI that replies to you

The replies are generated by DeepSeek, an AI provider, not by a model we run ourselves. To generate a reply, the app’s server sends DeepSeek the messages of the current reflection and a short description of you assembled from your profile: preferred name, pronouns, age bracket, focus areas, communication style, whether you have been in therapy before, and your language.

Your email address and account identifier are never sent. So that Maxi can pick up where you left off, every message also carries a short digest of your recent history: the summaries of up to three reflections from the last 30 days and the feelings you confirmed after them, up to two topics you asked to come back to, the moods from your check-ins in the past week (without the notes attached to them), and up to three journal notes, shortened to 400 characters. A journal note is only ever sent if you switched on “Let Maxi use this” for it — the switch is off on every new note. A reflection you start from a note sends that note along with it. The digest of recent reflections can be switched off, under Settings → Personalisation → Memory, as “Use recent reflections for context”. With it off, no summary of an earlier reflection and none of the feelings you confirmed after one are sent, from Maxi’s next reply onwards. The topics you asked to come back to, your check-in moods and the journal notes you permitted one by one are not affected by that switch.

The same applies when a reflection ends: the conversation is sent once more so a summary and its themes can be written, and individual messages are sent to a safety classifier that looks for signs of crisis.

A Future report, if you ask for one, is computed on our server from your own check-ins, finished reflections and the journal notes you let Maxi use — the numbers never come from the AI. To word each section, the server sends DeepSeek those numbers with the counts behind them, your language, and — only if “Use recent reflections for context” is on — the summaries of up to twelve recent reflections, shortened to 280 characters, plus the permitted notes. With that switch off, no reflection summary leaves the server for a report. A conversation you open about a report sends the report’s numbers and wording along with each message, under the same rules as a reflection.

DeepSeek is based in China, and processing happens on its infrastructure. That means conversation text leaves the European Economic Area and is handled under DeepSeek’s own terms and privacy policy as well as this one. If that is not acceptable to you, this app is not a good fit — there is currently no way to use the reflection feature without it.

When DeepSeek cannot answer — it is down, overloaded, or does not start replying in time — the same request is sent to Groq instead, so that you still get an answer. Groq then receives exactly what DeepSeek would have received, as described above, and nothing more; a reply that has already started appearing on your screen is never handed over midway. Groq is based in the United States, so in that case the text may be processed outside the European Economic Area too, under Groq’s own terms and privacy policy as well as this one.

Before your first conversation, the app asks for your explicit consent to this: what you write to Maxi, which may include information about your health, is sent to DeepSeek in China to generate the replies. That consent is the legal basis for sending it (GDPR Art. 6(1)(a) and Art. 9(2)(a)), and the app records when you gave it. You can withdraw it at any time in Settings → Privacy & data; from then on nothing you write is sent to an AI provider until you give it again. Withdrawing does not affect what was sent before, and your journal, check-ins and saved reflections keep working.

Memberships and payments

Memberships are sold through Apple’s App Store on iPhone and through Google Play on Android. The store takes the payment under its own terms, and Maxi never sees your card or payment details.

To know which plan you have, we use RevenueCat, a subscription service. RevenueCat receives your account identifier — a random string, never your email address — and the store’s record of your purchases (Apple’s or Google’s): the product, its dates, the price and currency, the storefront country, and whether it was a trial, a renewal or a refund. It receives nothing you write, and no mood, topic or reflection.

RevenueCat only hears about you when you open the membership screen, buy or restore a membership, already have one, or delete your account. RevenueCat is based in the United States, so this record is processed outside the European Economic Area, under its data processing terms.

When you delete your account, we also ask RevenueCat to erase any record it holds about you. Deleting your account does not cancel an App Store or Google Play subscription — only you can, in your Apple ID settings or in Google Play → Payments & subscriptions.

Where it is kept, and how it is protected

Data is stored in a Postgres database (Neon) and served by our API (Vercel).

Reflection content — conversations, summaries, themes, mood notes, journal notes and your preferred name — is encrypted by the server with AES-256-GCM before it is written to the database, so the stored rows are unreadable without a key that is not kept in the database. Everything travels over TLS.

On your device you can require Face ID, Touch ID or your passcode to open the app, and the app hides its contents in the app switcher.

Exports are rendered on your device and shared by you — the server never writes an export file that outlives the request.

How long it is kept

Reflections, journal notes and mood check-ins are kept until you delete them or delete your account.

A deleted reflection disappears from the app at once, and both it and the conversation it came from are erased from the database within seven days.

A reflection you started but never finished is closed automatically after 24 hours and erased within seven days after that.

Reports you send about a reply are kept until you delete your account. Deleting the reflection does not delete the report, but the report holds no text of the conversation.

A Future report is kept until you delete it or delete your account. Deleting one erases its contents at once; a record that it was written stays, because it counts towards the one report allowed per billing period.

Deleting your account deletes your sign-in account and every row belonging to you immediately.

Your membership record is kept while your account exists and is deleted with it.

Your reply count and the token and cost records are kept while your account exists and are deleted with it.

Your rights

You can export everything the app holds about you, in readable form, from Settings — that is your right of access.

You can correct your profile at any time, and delete all of your data and your account from Settings. Deletion is immediate and cannot be undone.

Under the GDPR you may also object to or ask us to restrict processing, and you may complain to the data protection authority in your country.

To exercise any right that the app does not already give you a button for, write to czechspeed1337@gmail.com.

Processors we rely on

• Clerk — accounts and sign-in.
• Neon — the database.
• Vercel — hosting for the API.
• DeepSeek — generating replies, summaries and the safety classification, as described above.
• Groq — the same work when DeepSeek cannot answer, as described above.
• Upstash — rate limiting.
• Sentry — crash and error reports.
• RevenueCat — membership status and App Store / Google Play purchase records, as described above.

Adults only

This app is for people aged 18 and over, and asks your age bracket during onboarding. It is not designed for children, and we do not knowingly hold data about them.

This is not medical care

Maxi is not therapy, medical advice, or an emergency service. It cannot diagnose, treat, or take responsibility for a crisis. If you are in danger, contact your local emergency number or a crisis line — the app lists them under Settings.

Changes to this policy

If this policy changes in a way that affects what happens to your data, the new version appears here with a new date, and the app tells you before the change takes effect.