1. Providers we use
| Provider | Purpose | Data | Where it is processed | Transfer safeguard |
|---|---|---|---|---|
| Hetzner Online GmbH (Germany) | Servers that run the Autopilot application, its PostgreSQL database, website and application logs, and the tools we run ourselves | All account data and all company content held in the database; data in transit to and from the application; logs; website analytics; records of email sent | Falkenstein, Germany | None needed: processed in the EU by a German company |
| Cloudflare, Inc. (United States): R2 object storage, DNS, email authentication reports | Storage of Library files and message attachments; storage of database backups; DNS for our domains; receiving reports about email sent in our name | Library files and attachments; daily database backups and the continuous log, kept 7 days; reports listing sending servers and domains, without message content | Chosen by Cloudflare today, and may be outside the EEA. We are moving these objects to R2 buckets under Cloudflare's EU jurisdiction | EU-US Data Privacy Framework, with the standard contractual clauses as fallback, under Cloudflare's data processing addendum |
| Vercel Inc. (United States): AI Gateway | Routing AI model requests for search indexing, result ordering, naming, Agent selection, and Agent work on managed AI | Text passages, images, search queries, prompts, context and model output, as described in the Privacy Policy, section 6. Vercel states that the gateway keeps no prompts or outputs after a request is completed | Global: any Vercel region | Standard contractual clauses (Commission decision 2021/914) under Vercel's data processing addendum. Vercel also states that it is certified under the EU-US Data Privacy Framework |
| Vercel Inc. (United States): Sandbox | Work Machines that we run for companies with managed compute | The repository working copy, Library files and session state an Agent works with | European Union | Stored in the EU. For access from outside the EEA: standard contractual clauses under Vercel's data processing addendum |
| Amazon Web Services EMEA SARL (Luxembourg): Amazon SES | Delivery of service email | Recipient email address; subject and content of the email; whether it was delivered, bounced or reported as spam | Frankfurt, Germany (region eu-central-1) | Processed in the EU under a contract with a Luxembourg company. For access from outside the EEA: standard contractual clauses, which are part of the AWS data processing addendum |
| Google Cloud EMEA Limited (Ireland): Google Workspace | Our own mailbox at questpie.com | Email that people send to us, and our replies | Any country where Google or its subprocessors have facilities | Contract with an Irish company. For onward transfers: Google's processor-to-processor standard contractual clauses, or the EU-US Data Privacy Framework where Google relies on it, under Google's Cloud data processing addendum |
2. Model providers behind the Vercel AI Gateway
The gateway passes each request to a company that runs the selected model. Vercel engages these companies as its own subprocessors under its data processing addendum, which carries the standard contractual clauses onward. Every request is routed only to providers that do not train models on it.
| Provider | Model and purpose | Data | Where it is processed | Transfer safeguard |
|---|---|---|---|---|
| Google (Gemini API) | Google's embedding model: search indexes for every company | Text passages from the Library, Tasks and messages; images from the Library | Chosen by Google; no EU region is offered for this model | Through Vercel, as stated above |
| DigitalOcean, serving the model "Jev" published by TypeSafe AI, Inc. | Ordering of search results for every company | The search query and up to 16 candidate passages, each cut to 1,000 characters | Chosen by DigitalOcean; no EU region is offered for this model | Through Vercel, as stated above |
| OpenAI, Microsoft (Azure) or Amazon Web Services (Bedrock), whichever serves the request | OpenAI's models: naming, Agent selection, and the default model for Agent work on managed AI | Short pieces of text for naming and selection; for managed AI, the Agent's prompts, context and output | Chosen per request; may be outside the EEA | Through Vercel, as stated above |
| Amazon Web Services (Amazon Bedrock), serving Claude models made by Anthropic, PBC | Claude models (Sonnet, Haiku): Agent work on managed AI, when a company chooses a Claude model for an Agent | The Agent's prompts, context and output | Chosen by Amazon Web Services per request; may be outside the EEA | Through Vercel, as stated above |
| Google (Gemini API or Google Cloud Vertex AI) | A small Google model used for naming and Agent selection only when the OpenAI model is unavailable | Short pieces of text | Chosen per request; may be outside the EEA | Through Vercel, as stated above |
A model provider may keep a request for a time under its own terms, for example to detect abuse. Zero-retention routing is not available for the embedding and ranking models.
3. From the start of paid plans
- Stripe and Sold through Link, LLC. Admitted accounts can test the workspace free during the closed beta until commercial launch with their own AI and computer, without a payment card. Managed AI and compute require a separate payment or an approved funded grant. From launch, subscriptions and credit are sold through Stripe Managed Payments. The merchant of record is Sold through Link, LLC, a Stripe company: it takes the payment on our behalf, works out, collects and remits VAT or sales tax, and sends the receipt and invoice. QUESTPIE remains the seller and the provider of the service. The payment partner receives the payment and billing data you give it at checkout and handles it under its own terms and privacy policy. It is listed here for completeness; it does not process company content.
4. Software we run ourselves
These are open-source tools installed on our own servers at Hetzner. They are not separate companies receiving data.
- Umami, for website analytics on questpie.com.
- Plunk, for composing and queueing service email, which it hands to Amazon SES for delivery. Service email carries no open or click tracking.
5. Services a company connects itself
These are not our subprocessors. A company chooses them and uses them under its own agreement with the provider:
- GitHub, when a company connects a repository to a Project.
- The company's own AI provider, when it runs Agents on its own computer with its own plan.
- Any tool connected over MCP.
6. Changes
Before a new subprocessor starts processing company content, we update this page and tell the Owners of each company by email at least 30 days ahead; where a provider has to be replaced urgently to keep the service running or secure, we tell them as early as we can. The date at the top shows the last change. To ask about a provider, to ask for a copy of a safeguard or to object to a change, write to privacy@questpie.com.