Subprocessors and where data goes
The complete list of third parties that process data on our behalf. If it is not on this page, it does not receive your data.
The list
| Provider | What it does | Data it sees | Where |
|---|---|---|---|
| Netcup GmbH | Hosting and infrastructure | Everything stored, at rest on their infrastructure | Austria (Vienna) |
| OpenAI | Writes the wording of each answer, and the summary on a handoff ticket | The question, the retrieved passages, and the conversation being summarised | United States |
| Voyage AI | Converts text into the numeric form used for searching | The question text, and your documentation when it is indexed | United States |
Your own helpdesk receives a handoff transcript when a visitor asks for a person. It is your provider under your agreement, not our subprocessor, and it is listed on your ticket configuration rather than here.
What each one does not see
Neither OpenAI nor Voyage AI receives your account credentials, your billing details, or your helpdesk credentials. Neither receives handoff names and email addresses, except where those appear in a transcript being summarised.
Netcup provides infrastructure and does not access application data in the ordinary course.
Training
Data sent to OpenAI and Voyage AI is processed under their API terms and is not used to train their models. We do not use customer conversations to train or fine-tune anything ourselves.
Transfers outside the EEA
OpenAI and Voyage AI process data in the United States. These transfers rely on the European Commission's standard contractual clauses, incorporated into our agreements with each provider.
If a US transfer is unacceptable to you, say so before you sign. It is not something we can configure away: generating an answer and searching by meaning both depend on these providers today, and pretending otherwise would be worse than telling you now.
Self-hosted search, deliberately
Our database and our search process run on our own servers rather than on a managed cloud search service. This is a deliberate architectural choice and the reason your documentation and conversation history are not sitting in a third-party multi-tenant cluster.
Changes to this list
We will update this page before a new subprocessor begins processing customer data, and we will notify customers on a Pro plan directly, as their agreement requires. If you want notice of every change regardless of plan, ask us and we will add you to that list.
Getting a data processing agreement
Email dimitar@usekaras.com. Our agreement names every provider above, includes the standard contractual clauses, and commits us to telling you about changes.