Legal
Subprocessors
Last updated 23 October 2026
These companies process customer data for Velkri, only to provide their service to us. Our privacy policy explains what data each kind of processing involves. We update this page and its date when the list changes, and email workspace owners at least 30 days before we add a new subprocessor. To object to a new one, write to [email protected] before it takes effect.
| Company | What it does for Velkri | Where |
|---|---|---|
| Hetzner Online GmbH | Servers that run Velkri: the app, database, search index, calls and meeting bots | Germany |
| Cloudflare, Inc. | DNS, content delivery and network protection; storage for files and recordings | Global network, so it may serve users from any country; files stored in the EU |
| Amazon Web Services, Inc. | Email delivery (Amazon SES): sign-in codes, invitations and notices | Germany |
| OpenRouter, Inc. | Routes AI requests to the model providers below | United States |
| AI model providers, through OpenRouter | Run the models for answers, summaries, captions, transcription, picture descriptions and search indexing | United States and EU |
| 650 Industries, Inc. (Expo) | Push notifications to the mobile apps, delivered through Apple and Google | United States |
| Paddle.com Market Ltd | Payments in US dollars, as merchant of record, when paid plans launch | United Kingdom |
| Razorpay Software Private Limited | Payments in Indian rupees, when paid plans launch | India |
AI model providers
OpenRouter sends the language models' requests only to providers that keep no copy and do not train on them, chosen from this list: DeepInfra, Together AI, Fireworks AI, Baseten, Parasail, Crusoe, CoreWeave, Nebius, Cloudflare, Novita AI, Groq, Cerebras, SambaNova, Amazon Bedrock and Google Vertex AI. Transcription runs on DeepInfra. Search indexing (embeddings) runs on OpenAI or Microsoft Azure, which do not train on what we send and may keep it for a short time to detect abuse.
Services we run ourselves
These run on our own servers at Hetzner, so they are not subprocessors: LiveKit for calls, Attendee for the meeting bot, PostgreSQL for the database, Qdrant for the search index, NATS and Centrifugo for messaging and live updates, Infisical for secrets, SearXNG for web search, Metabase for internal usage reports, and our document extractor.
Services you connect
Services a workspace chooses to connect, such as Slack, Google Calendar, Jira or GitHub, are not our subprocessors: the customer chooses them, and they handle data under their own terms. A data processing agreement is available on request from [email protected].
Earlier versions
None yet: this list was first published on 23 October 2026.