privacy policy

who we are

Matraca provides bilingual keyboards and practice features for Android and iOS, plus a macOS voice and practice app. This policy explains how we handle your personal data when you use the apps, the keyboard, and our website.

The entity behind Matraca is based in São Paulo (Brazil) and Lisbon (Portugal) and acts as the controller of the data described here. For anything privacy-related, reach us at contato@matraca.ai.

keyboards on Android and iOS

On iOS, to translate while you type, the Matraca keyboard needs the "Full Access" permission that iOS requests when you install a third-party keyboard. We know that permission is scary, and it deserves an honest explanation. The macOS app does not use this permission.

On Android, you enable and select the keyboard in system settings. When you tap to convert, the selected text or current line is sent to Matraca’s backend for translation, corrections, and storage in your account history. Ordinary typing outside a requested conversion is not sent to our service. You can switch keyboards or disable Matraca in Android settings.

When you grant Full Access and you're typing inside a Matraca translation field, the text you write is processed to produce the translation — partly on the device itself, and partly in our translation infrastructure when the language requires it. That processing exists for one reason only: to hand you the sentence back in the other language.

What we do NOT do, and never will:

• We do not log or store what you type in password fields, credit card fields, or any field iOS marks as secure — the keyboard is disabled in those contexts.
• We are not a keylogger: we do not silently capture everything you type in other apps outside of Matraca's translation flow.
• We do not sell, rent, or trade what you type. Your keystrokes are not a product.
• We do not use the content of your messages for targeted advertising.

Text sent for translation always travels encrypted and is used only to produce the response and run app features (like your practice history and streak). You can revoke Full Access at any time in iOS Settings — without it, the keyboard's translation features stop working.

what we collect

We collect the minimum needed to make the app work and improve it:

• Account data: account identifier, display name, email, and authentication data from your sign-in provider. If you provide or import them, we also process your profile photo, biography, and date of birth.
• Text processed for translation: the sentences you write in the translation flow, and their translations, tied to your account to power history and practice.
• Audio and transcription: when you voluntarily start a voice feature, we capture your recording and generate a text version to provide the requested feature.
• Contacts and invitations on Android: with your permission, we read names and email addresses from your contacts on the recipient-selection screen. Only the email addresses you select or enter are sent to the backend when you confirm invitations; we do not upload your entire address book.
• Notifications on Android: installation identifier, notification token, language, time zone, and notification preference, to register the device and deliver service notices.
• Reports and blocks on Android: the reason and details you submit, identifiers for the report, your account, and the reported group, member, or exercise. We also store the users you block to apply your preference.
• Streak and social data: consecutive days, completed chapters, friend connections, and the friendly pressure that keeps the conversation going.
• Usage analytics: how you move through the app, which features you use, and error reports — aggregated wherever possible.
• Device data: model, Android, iOS, or macOS version, app version, language, time zone, technical identifiers, and diagnostics. On Android, Firebase Remote Config also determines the approximate country from the IP address to provide app configuration, without requesting GPS location access.
• Free-access enforcement and fraud prevention: allowance usage and a pseudonymous identifier derived from the Apple or Google account, without storing the provider's raw identifier.

audio, microphone, and transcription

Matraca captures audio only when you voluntarily start a voice feature and while recording is active. Android, iOS, or macOS microphone permission is requested before the first capture.

In the macOS app, recording and transcription are processed locally on your Mac with the Parakeet model. Raw audio is not sent to our backend or to the providers below. Recordings may remain in local history according to the retention setting you choose and can be deleted in the app. Resulting text is sent only when needed for features you request, such as translation, exercises, account history, or feedback.

On Android, voice transcription is remote: the app sends the recorded audio to our backend. In this flow and in iOS voice features that use remote transcription, our backend validates the account and request, then sends the audio to one of the providers below. The purpose is to turn your speech into text to fill in your answer, generate the requested translation, and, when offered by the feature, provide learning feedback. We do not intentionally store raw audio in our database or log its contents.

• AssemblyAI — receives the audio to produce the transcript. As soon as a transcript is completed, or fails after being created, we send an API request to delete the transcript artifact and its associated audio file. If immediate disposal cannot be completed, AssemblyAI states that automatic deletion of uploaded audio begins after 24 hours and completes within 48 hours; limited metadata may be retained for logging and billing.
• Deepgram — receives the audio to produce the transcript. Every Matraca request uses the mip_opt_out option, which excludes the content from its model-improvement program; according to Deepgram, data from these requests is retained only for the time needed to process them.
• OpenAI — does not receive raw audio. When needed for translation, language conversion, or learning feedback, it may receive only the transcript and the languages involved. We send these requests with application-state storage disabled (store=false). OpenAI does not use API data to train its models by default, but may retain content in abuse-prevention logs for up to 30 days, unless a longer period is required by law or reasonably needed for safety.

You can revoke microphone access at any time in Android app permissions or in iOS or macOS Privacy & Security settings. This prevents new recordings and disables features that rely on voice without affecting features that do not use the microphone. On macOS, you can also delete local history in the app. To request deletion of a transcript or other data tied to your account, email contato@matraca.ai.

how we use your data

We use your data to:

• Operate the keyboard and deliver translations while you type.
• Transcribe audio you choose to record and provide the requested voice features.
• Maintain your account, streak, chapters, and connections.
• Process subscriptions and payments.
• Improve the product, fix bugs, and understand what works.
• Send important service notices and, with your consent, product news.
• Comply with legal obligations and protect against fraud and abuse.

legal bases

We process your data on the basis of contract performance (delivering the app and feature you requested), your consent (where applicable, including microphone access), our legitimate interests (security, product improvement, fraud prevention), and compliance with legal obligations. Under LGPD and GDPR, you can withdraw consent at any time.

third parties and processors

We don't work alone. To run Matraca, we share strictly necessary data with providers acting as processors, under contract and confidentiality obligations:

• Stripe — payment and subscription processing.
• Resend — delivery of email invitations you request; receives the recipient address and invitation details to deliver the message.
• RevenueCat — purchase validation and subscription-status management on Android and iOS; RevenueCat receives technical account identifiers and purchase events, but not full card details.
• Intercom — customer support and messaging when you open that feature in the app; the technical account identifier, device data, and conversation content may be shared.
• Google Cloud and Firebase — authentication, backend, database, photo storage, and service hosting. On Android, we also use Play Integrity/App Check to verify app integrity, Firebase Cloud Messaging for notifications, and Firebase Crashlytics for crash reports and diagnostics. These services receive technical identifiers and data needed for those functions.
• AssemblyAI and Deepgram — remote transcription of audio voluntarily submitted through Android and iOS voice features; audio from the macOS app is not sent to these providers.
• OpenAI — processing of text and, where applicable, transcripts for translation, language conversion, and learning feedback; raw audio is not sent.
• Analytics and error-reporting providers — to understand usage and stability.
• Google Play — Android app distribution, purchase processing, and management of subscriptions purchased through the store.
• Apple — App Store distribution and billing on iOS; on macOS, Developer ID signing, notarization, and DeviceCheck integrity verification, without Mac App Store distribution.

Tutoring also uses Inngest to process background tasks. In this flow, we send the account identifier, practice reference and time, its source (speech or typing), languages, time zone, and the original and converted text. Raw audio is not sent to Inngest. Cleaning up history in Firebase does not automatically delete copies of these data in the tutoring queue.

We do not sell your personal data to anyone.

reports and blocks

You can report groups, members, and exercises within the Android app. We use the submitted reason, details, and references to review inappropriate content and prevent abuse. To preserve the report's context, we may store the member's display name or the group's name and rule. Exercise reports store references to the set and item, without duplicating exercise text or audio in the report. The team responsible for the service may access this information to handle the case.

Blocks are personal and tied to your account. We store the blocked person's identifier and display name to hide them and the groups they own, prevent invitations between you, and let you manage your blocks. You can unblock people in the Blocked users list. Reports and blocks follow the retention and deletion rules described in this policy.

payments and subscriptions

Payments for subscriptions purchased on the website are processed by Stripe. Your card details are collected and handled directly by Stripe under PCI-DSS standards — Matraca does not store your full card number. We receive from Stripe only what we need to manage your subscription, such as payment status and the last digits of your card. Subscriptions made through the App Store are managed by Apple. On Android, in-app purchases are processed by Google Play. Matraca and RevenueCat receive purchase identifiers and status needed to enable access, without receiving full card details used in the store.

data retention

On macOS, raw audio remains in local history according to the retention policy configured in the app and is not kept in Matraca's database. On Android and for iOS features with remote transcription, audio exists temporarily on the device and in backend memory during processing and is discarded after the request. With AssemblyAI, we request deletion of the transcript and associated audio immediately after processing; in the contingency described above, uploaded audio is automatically deleted within 48 hours. With Deepgram, mip_opt_out requests are retained only during processing. With OpenAI, we use store=false, but abuse-prevention logs may contain the transcript for up to 30 days, subject to legal or safety exceptions.

The transcript returned to the app may become part of your answer or practice history when that is part of the feature you used. In that case, it is handled as account text and stored while the account is active or until you delete the content or request its deletion. When you delete your account or submit a request to contato@matraca.ai, we remove or anonymize linked data, except for what we must retain for legal, tax, security, or fraud-prevention obligations.

To prevent repeated account deletion and recreation from resetting the free allowance, we retain a pseudonymous access ledger for up to 730 days after its most recent use. It contains only accumulated usage, technical versions, and creation, use, expiry, and account-deletion dates. It does not contain a Firebase UID, email, name, or the raw Apple or Google account identifier. Each new use renews this period; the record is not used for advertising.

delete your Matraca account

On Android, you can request deletion in the app under profile > settings > delete account and confirm your Google identity. The request covers the account and linked personal data, subject to the exceptions described in the data-retention section.

You can also request deletion without accessing the app: email contato@matraca.ai with the subject Delete Matraca account and include the email address of the account you want to delete. You can use the same channel to request deletion of specific data without deleting the entire account.

To stop charges for a subscription purchased through Google Play, cancel it in the subscriptions section of your Google Play account.

security

We use encryption in transit and at rest, access controls, and security practices proportionate to the risk to protect your data. No system is perfect, but we treat your data the way we'd want ours treated. If a significant incident occurs, we will notify you and the relevant authorities as required by law.

international transfers

Matraca operates between Brazil and the European Union, and our providers may process data in other countries. When we transfer data internationally, we apply the safeguards required by LGPD and GDPR, such as standard contractual clauses and adequacy checks.

your rights

Under LGPD and GDPR, you have the right to access, correct, update, port, and delete your data, as well as to object to or restrict certain processing and withdraw consent. To exercise any of these, write to contato@matraca.ai — we respond within the legal timeframes. You may also lodge a complaint with the competent data protection authority (ANPD in Brazil, or your EU supervisory authority).

children

Matraca is not made specifically for children, and we do not knowingly collect children's data without the consent required by law. If we learn we've collected data from someone below the permitted age without the necessary authorization, we'll delete or correct that data.

changes to this policy

This policy may change as the product evolves or the law changes. When a change is significant, we'll let you know in the app or by email and update the date at the top. Continuing to use Matraca after that means you agree to the current version.

talk to us

Question, request, or complaint about privacy? Send a message to contato@matraca.ai. We're in São Paulo and Lisbon.