Subprocessors

Last updated: October 18, 2026

This page lists the third parties that process personal data for Tjati, what each one does, what data it receives and where that happens. It supports the Privacy Policy, which controls if the two differ.

How to read this list

A provider is listed here if it processes personal data of Tjati members or website visitors on our behalf. Locations come from each provider's own public documentation; where a provider publishes none, the table says so. What each provider keeps and for how long depends on its own terms. See the Privacy Policy for how we use these providers.

Infrastructure and hosting

Infrastructure and hosting providers

Provider	Purpose	Data involved	Location
Hetzner	Servers that run Tjati, including its database, cache and background job queues.	All account and Service data: account details, your content, conversations, and operational logs.	Germany and Finland (European Union).
Cloudflare	Object storage for uploaded documents, companion images and database backups; delivery and protection of the website and app traffic; the waitlist database and its bot check.	Files you upload, generated companion images, backups, request metadata such as IP address, and waitlist email addresses.	Backups and the waitlist database are restricted to the European Union. Other traffic passes through Cloudflare’s global network.

Accounts, subscriptions and notifications

Account, subscription and notification providers

Provider	Purpose	Data involved	Location
Clerk	Sign-in and account identity.	Email address, name, avatar, sign-in identifiers and session data.	United States and elsewhere; Clerk states it may process data anywhere in the world.
Apple	Sign in with Apple, App Store purchases and subscriptions, and delivery of push notifications to iPhones.	Apple account identifiers you choose to share, purchase records, and notification delivery data.	Worldwide; Apple states personal data is generally stored in the United States.
Google (Google Play and Firebase Cloud Messaging)	Where the Android app is available: Google Play purchases and subscriptions, and delivery of push notifications to Android devices.	Purchase records and entitlement identifiers, push notification token and the text of the notification.	Worldwide, per Google’s privacy policy.
RevenueCat	Subscription and entitlement management, including distinguishing customer, internal-testing and Apple-review accounts.	Store transaction and entitlement identifiers, product and subscription status, an app-user identifier, your verified email address, and non-sensitive account classification and environment labels.	United States.
Expo	Push notification delivery to iOS and Android devices; Expo hands notifications to Apple and, for Android, Google’s Firebase Cloud Messaging.	Push notification token and the text of the notification.	United States.

Email

Email delivery provider

Provider	Purpose	Data involved	Location
Resend	Delivery of the emails Tjati sends, and delivery reports back to us.	Recipient email address, the message, and delivery events.	Emails are sent from Ireland (European Union). Resend states that its account data and logs are stored in the United States.

Analytics and diagnostics

Analytics and diagnostics providers

Provider	Purpose	Data involved	Location
PostHog	Product and website analytics.	In the app: product events, device details and your account identifier. On tjati.com: anonymous pageviews, browser and device details, and referring domain; no account or waitlist identity, persistent browser storage or session replay.	Frankfurt, Germany (European Union).
Sentry	Crash and performance diagnostics.	Crash reports, device and network details and your account identifier; no name or email.	Frankfurt, Germany (European Union). Sentry keeps its own account and organization data in the United States.
Langfuse	Monitoring and evaluation of AI requests, and storage of AI prompt templates.	Metadata about AI requests, such as model, timing, token counts and identifiers. In the current production configuration it does not receive prompts or replies.	Ireland (European Union).

AI language models and search

These providers receive your content when you use an AI feature. Which model handles a given request can change as the Service evolves; every provider that does so appears here.

AI language model, embedding and reranking providers

Provider	Purpose	Data involved	Location
OpenAI	Language models that understand requests, write replies and decide actions; image generation for your companion.	Your prompts and the context needed to answer them, such as selected items, notes, memories, retrieved document passages and tool results; the companion appearance description you enter.	Processing is not restricted to one region (provider default).
Anthropic	Language models, available as an alternative provider.	The same categories as other language-model providers, if requests are routed to it.	Stored in the United States; Anthropic states processing may be routed to the United States, Europe, Asia and Australia.
Google	Language models (including as a fallback) and embeddings for search over your documents and items.	Prompts and context, and text extracted from your documents and items for embedding.	Any country where Google or its agents maintain facilities, per Google’s Gemini API terms.
xAI	Language models for understanding and reasoning about requests.	Your prompts and the context needed to answer them.	Not published by the provider; treated as outside the European Union.
Cohere	Reranking of search results.	Your search query and the candidate passages being ranked.	United States, Canada and the United Kingdom, per Cohere’s privacy policy.
TypeSafe	Short judgments about how to interpret and route a request.	The text of your request and the minimal context needed to classify it.	Not published by the provider; treated as outside the European Union.

Voice

Voice providers

Provider	Purpose	Data involved	Location
Groq	Speech-to-text for voice requests (English, and as a fallback for streaming).	Audio of what you say while a voice session is open. Tjati does not keep a stored recording.	United States, with processing operations in other countries, per Groq’s privacy policy.
Deepgram	Streaming speech-to-text for Arabic.	Audio of what you say while a voice session is open.	United States.
Cartesia	Text-to-speech for spoken replies.	The text of the reply being spoken.	Not published by the provider; treated as outside the European Union.

Accounts you connect

When you connect an external account such as Gmail or Google Calendar, you direct Tjati to read from that service. The provider is not our subprocessor; it acts on your instruction and under its own terms. See the Privacy Policy.

Questions and updates

Write to us to ask for this list, to ask about a provider, or to raise a concern. Put Subprocessors in the subject line.

Tjati privacy and support

support@tjati.com