Sub-processors
Last updated 2026-11-16
We use the following sub-processors to deliver the undocsend service. We update this list as our stack evolves and notify Customers at least 30 days before adding a new sub-processor. Customers who object to a new sub-processor may terminate the Service in accordance with our DPA.
| Vendor | Purpose | Data shared | Region | DPA |
|---|---|---|---|---|
| Supabase | Application database and authentication | All app data, user accounts, session tokens | US | Supabase DPA |
| Cloudflare R2 | Document storage | Uploaded documents, generated previews and thumbnails | Global | Cloudflare DPA |
| Vercel | Application hosting; Vercel AI Gateway, which routes AI requests to the AI providers below | Application traffic, request logs; AI prompts and outputs in transit (not retained by Vercel) | Global | Vercel DPA |
| Resend | Transactional email delivery | Recipient email addresses and email content (magic links, notifications) | US | Resend DPA |
| Anthropic | AI model inference via Vercel AI Gateway, zero data retention requested | Prompts built from document content and visitor questions, and the generated output | US | Anthropic Privacy Policy |
| OpenAI | AI model inference via Vercel AI Gateway, zero data retention requested | Prompts built from document content and visitor questions, and the generated output; text sent for search embeddings | US | OpenAI Data Processing Addendum |
| Amazon Web Services (Amazon Bedrock; Claude on AWS) | AI model inference via Vercel AI Gateway, zero data retention requested | Prompts built from document content and visitor questions, and the generated output | US | AWS Data Processing Addendum |
| Google Cloud (Vertex AI) | AI model inference via Vercel AI Gateway, zero data retention requested | Prompts built from document content and visitor questions, and the generated output | US | Google Cloud Data Processing Addendum |
| Microsoft (Azure OpenAI) | AI model inference via Vercel AI Gateway, zero data retention requested | Prompts built from document content and visitor questions, and the generated output | US | Microsoft Products and Services DPA |
| xAI | AI model inference via Vercel AI Gateway, zero data retention requested | Prompts built from document content and visitor questions, and the generated output | US | xAI Data Processing Addendum |
| PostHog | Product analytics and feature flags | Page views, custom events, IP-derived geolocation | US (EU region available) | PostHog DPA |
| Sentry | Error monitoring | Stack traces, browser metadata, redacted request context | US | Sentry DPA |
| Stripe | Payment processing and billing | Account email, billing details, card details (entered directly with Stripe), payment metadata | US | Stripe DPA |
AI sub-processors and BYOK
When a Customer enables AI features, AI requests are served by the AI providers listed above (Anthropic, OpenAI, Amazon Web Services, Google Cloud, Microsoft and xAI), reached through Vercel AI Gateway, which chooses the provider for each request. Customers may bypass these defaults by configuring Bring Your Own Key (BYOK) for a supported provider.
AI requests paid for with undocsend AI credits are sent through Vercel AI Gateway with zero data retention required on every request, so they are served only by AI providers that have a zero-data-retention agreement with Vercel. Some providers keep limited data for abuse monitoring under those terms, and providers may cache prompts for short periods.
A BYOK request is sent through Vercel AI Gateway using the Customer’s own key and is governed by the Customer’s agreement with that provider, not by the zero-data-retention routing above. If the Customer’s key fails, the gateway may complete the request on undocsend’s own provider account, without zero data retention.
Notifications
When we add a new sub-processor, we publish the update here and email Customers at least 30 days before the new sub-processor begins processing data. To receive these notifications, ensure your workspace billing-contact email is up to date.