Kakikomi

Privacy Policy
Auf Deutsch lesen

Kakikomi collects nothing about you. There is no account, no analytics, no advertising and no tracking of any kind. The dictionary, the handwriting recognition and the flashcards all work on your device, without a network. Text leaves the device only for online translation and for Sensei — the grammar explanations and the chat — and only when you use them. “Always offline” in the settings keeps translation on the device.

Last updated: 14 October 2026

Who is responsible

Mael Gerhard
Augasse 8, 79798 Jestetten, Germany
Contact: hello@kakikomi.app

What stays on your device

All of the following is stored only on your device and is never transmitted anywhere:

Kakikomi does not read anything from the apps it appears on top of. The floating bubble draws its own window and has no access to the content behind it. Text reaches the app only when you select it yourself and choose Kakikomi from the selection menu, when you type or paste it in, or when you use “Read screen”.

Read screen (part of Pro) takes a single screenshot — only after you agree in Android's own “Start recording” dialog, every time. Text recognition runs on your device; the image is never transmitted or stored and is discarded as soon as you have picked the lines you want. The lines you pick are then handled like any other text you enter.

When data does leave your device

Only in the situations listed here, and never without an action on your part.

WhatTo whomWhy
The text you translate Our server (api.kakikomi.app, Hetzner Online GmbH, Germany), and from there OpenAI, L.L.C., USA With online translation switched on (the default), while you have a Sensei subscription or free Sensei actions left. The text and the two languages go to our server, which has OpenAI's language model translate it. Our server keeps a one-way fingerprint of the text and the translation for up to 90 days, so the same text does not need to be sent again; it does not keep the text itself. OpenAI processes the request on our behalf under the terms described below. Set “Always offline” and this never happens.
The text you entered for translation DeepL SE, Germany Only if you have stored your own DeepL API key, left online translation enabled, and the translation above is not available or failed. Every result states how it was translated — by Sensei, by DeepL or on your device. Set “Always offline” in settings and this never happens. See DeepL's privacy policy.
A model download request Google LLC The handwriting model, the offline translation models and the Japanese text recognition model (through Google Play services) are downloaded once. Google receives the usual request data, including your IP address. No content of yours is sent — recognition and translation then run on your device.
The word to be spoken aloud Your device's speech engine Only when you tap the speaker icon. Kakikomi prefers a voice that works offline. If your device only has an online voice, that word is passed to the engine's provider — usually Google.
A sentence you ask Sensei to explain, with the words Kakikomi found in it Our server (api.kakikomi.app, hosted by Hetzner Online GmbH, Germany), and from there OpenAI, L.L.C., USA Only when you tap “Explain the grammar”. The sentence, its on-device word analysis and your chosen language go to our server, which asks OpenAI's language model for the explanation. No account or device ID is sent; with a subscription, see the subscription row below.

Our server does not store the sentence. It keeps a one-way fingerprint of it, the generated explanation for up to 90 days so the same sentence does not need to be sent again, and a daily counter. OpenAI processes the request on our behalf; under its API terms it does not use it to train models and may keep it for up to 30 days to detect abuse. The transfer to the USA is covered by OpenAI's data processing terms, including the EU standard contractual clauses.
Your messages in a chat with Sensei, and what the chat is about Our server, and from there OpenAI, L.L.C., USA Only when you send a message in the Sensei tab. With each message, the app sends the recent part of that conversation (at most 12 messages), the sentence it is about if there is one, your level and language. For card suggestions it also sends up to 40 words you already have as cards, so they are not suggested again; for a quiz, the words, readings and meanings of up to 10 of your upcoming cards.

Our server stores none of it: it passes the request on and forgets it. Conversations are kept only on your device, in your grammar notebook. OpenAI processes the request on our behalf under the same terms as above.
An explanation you report Our server Only when you tap “Report” under an explanation. The sentence and the explanation are stored so that we can check it, and deleted once checked. The explanation is no longer given to anyone else.
The e-mail address you enter to join the beta, and your device type (iPhone/iPad or Android) Our server, and from there our own mailbox (Proton AG, Switzerland) Only if you enter it on this website to join the beta. We store the address, the device type, the page language and, if the link you came from carried one, a short channel tag such as “ig” (no cookie, no tracking). We e-mail it to ourselves and send a confirmation with your spot to the address you entered. When we let you into the beta, a second e-mail tells you how to join (Google Play or TestFlight); for Google Play we need the address of your Google account, which you can reply with. For Android you then get a licence key; we note which key went to your address. We record when each step of your beta happened (signup, our e-mails, joining Google Play, the first time your key is redeemed) so that we can ask after about ten days how it is going. How you use the app is not recorded. After that we write only about the beta and the launch. We delete the address when the beta ends, or sooner if you ask us to: every e-mail carries an unsubscribe link that deletes it at once, and a reply works too.
Your answers to the five questions after you join the beta Our server Only if you answer them; every question is optional. We store your answers (your level of Japanese, what you read, what stops you, what you would pay for and your wish) together with the address you entered, and use them only to improve Kakikomi and to plan the launch. They are deleted together with the address.
Purchase details Google LLC Only if you buy Kakikomi Pro or the Sensei subscription. The purchase is handled entirely by Google Play; Kakikomi never sees your payment data and only asks Google whether the purchase exists. See Google's privacy policy.
Your subscription's purchase token Our server, and from there Google LLC Only with a Sensei subscription. Each Sensei request carries the token Google Play issued for your subscription, so our server can ask Google whether it is active. Our server keeps a one-way fingerprint of the token, the end date of the subscription and a monthly usage counter (a limit against automated misuse) — no name, no e-mail address, nothing about what you asked.
Your licence key and an installation ID Our server Only if you redeem a licence key from a language school, a teacher or the beta. The app sends the key and a random installation ID — not a device ID, not an account; it is made from chance on first use and disappears with the app — so our server can check that the licence is valid and count the devices a key runs on. Our server keeps a one-way fingerprint of the key, a fingerprint of the installation ID, the dates of the licence and a monthly usage counter. No name, no e-mail address, nothing about what you asked. Without a licence key, the installation ID is never sent.

On iPhone and iPad

The iOS app works the same way, with these additions:

This website and our server

This website and api.kakikomi.app run on a server of Hetzner Online GmbH in Germany. Like every web server, it records technical access data — IP address, time and requested address — in a log that rotates automatically and keeps only the most recent entries (at most 10 MB per service, which at current traffic is a few days). It is used only to keep the service running and secure. The website loads nothing from other servers and sets no cookies.

Legal basis

Where the GDPR applies, processing rests on Article 6(1)(b) — performing the service you asked for. Online translation, Sensei explanations and chat, reports, joining the beta, read screen and text-to-speech happen only on your explicit instruction; you can avoid all of them and still use the app. Checking a subscription also rests on Article 6(1)(b). The server log rests on Article 6(1)(f) — our legitimate interest in a service that runs and stays secure.

Your rights

Kakikomi holds no data that identifies you, except the address you enter to join the beta. Everything the app stores is on your device and disappears when you uninstall it or clear its data; single conversations can be deleted in the grammar notebook. The only other thing on our server that relates to a person is the fingerprint of a subscription token, which we cannot link to you; it expires with the subscription. For data held by Google, OpenAI or DeepL, please contact them directly.

Children

Kakikomi is not directed at children under 13 and collects no data that could identify anyone.

Changes

If this policy changes, the date above changes with it. Material changes will also be noted in the app's release notes.

Dictionary data

Kakikomi bundles KANJIDIC2 and JMdict by the Electronic Dictionary Research and Development Group, used under the Creative Commons Attribution-ShareAlike 4.0 licence. Word segmentation uses Kuromoji (Apache 2.0).