1. How we use subprocessors
We use a small number of providers to host and operate Chatrevia. Each processes data only to provide its service to us. Before a new subprocessor begins processing customer personal data, we update this page and email workspace owners at least 14 days in advance, as described in our Data processing agreement.
2. Infrastructure and operations
| Provider | Purpose | Data | Location |
|---|---|---|---|
| Hetzner Online GmbH | Servers that run the application, its scheduled jobs and its PostgreSQL database. | Account, workspace, conversation and visitor data. | Germany (Falkenstein) |
| Cloudflare, Inc. | DNS, content delivery, TLS and abuse protection; private storage for attachments (R2). | All service data in transit; attachments. | Global edge network; attachments in the European Union |
| OpenRouter, Inc. | Routes AI requests for answers, knowledge indexing, voice transcription and speech to the model providers below. | Visitor messages and recent conversation, relevant published knowledge, voice messages. | United States; model providers may process elsewhere |
| LiveKit, Inc. | Real-time audio and video for calls in the messenger, and hosting of the AI call agent. Only for workspaces that use calls. | Call audio and, while a visitor shares it, camera or screen video, in transit (not recorded); call participants’ display names. | United States (AI agent in US East); calls are routed through LiveKit’s nearest region |
| Google LLC (Gemini API) | Speech-to-speech AI for calls with the AI agent (Gemini Live). Only for workspaces that use calls. | Call audio, camera or screen video while shared, the agent’s instructions and relevant published knowledge; returns speech and transcripts. | United States and other Google locations |
| Stripe, Inc. and its affiliates | Subscription billing and payment processing. | Billing contact, workspace identifier, payment details collected directly by Stripe. | United States and other locations |
| Resend (Plus Five Five, Inc.) | Transactional email, such as verification, invitations and notifications. | Recipient names and email addresses, email content. | United States |
| Browser push services (Apple, Google, Mozilla, Microsoft) | Deliver push notifications to devices that turned them on. | Encrypted notification content and the device’s push address. | Varies by browser vendor |
3. AI model providers
Through OpenRouter, Chatrevia currently uses open-weight models from NVIDIA (Nemotron) and Alibaba (Qwen) to write answers, OpenAI models to index knowledge and transcribe voice messages, and MiniMax to read answers aloud. Which inference provider serves a request can vary.
Calls with the AI agent use Google’s Gemini Live models directly through the Gemini API, not through OpenRouter.
Some of these models are offered through free endpoints whose providers may log prompts and responses and use them to improve their services. Businesses should not publish knowledge they consider confidential, and visitors should not share sensitive information in a chat. We will update this section when the models we use change.
4. Integrations you choose
Workspaces can connect third-party tools, including Meta (WhatsApp and Messenger), Slack, Calendly, Cal.com, Shopify, Zoho, Klaviyo, HubSpot, Salesforce, Pipedrive, Attio, Close and Folk, as well as custom webhooks. These are not our subprocessors: the workspace decides what to connect and what data to send, and each tool is governed by the workspace’s own agreement with its provider.
5. Objections and questions
Customers may object to a new subprocessor under the Data processing agreement. Questions about this list: support@chatrevia.com.