Subprocessors
Last updated 2026-10-08. This is a plain-language draft that has not been reviewed by counsel and is not yet in force.
Who handles your data for us
These vendors process data on our behalf to run Meynote. Each one handles data only to provide its service to us. We are putting data processing agreements in place with them, so treat this page as a draft until that is done.
We update this page before a new vendor starts handling your data.
In use now
| Name | Purpose | Data it receives | Where |
|---|---|---|---|
| Railway | Hosts the website, the API and the background workers. | Whatever the service handles while it runs: account details, deck content, text and photos in transit and in memory, and server logs. | United States (US West) |
| Supabase | Managed Postgres database. | Account records, deck content, recipient addresses and notes (for 30 days), credit and billing records, consent and audit records. | United States (AWS us-west-2, Oregon) |
| Cloudflare | DNS, content delivery, web application firewall, and R2 storage for photos and published images. It also scans the published image host for child sexual abuse material. | Photos and published images, plus visitors' IP addresses and request details. | R2 storage in Western North America; delivery from Cloudflare's global network |
| Clerk | Sign-in, account profile and email verification. | Name, email address, sign-in credentials and sign-in activity. | United States |
| Stripe | Payments and billing. | Name, email address, billing details, the payment method (card numbers go to Stripe, not to us), plan and invoice history. | United States and other regions where Stripe operates |
| Resend | Delivers the deck emails you send, and our service notices. | The recipient's email address, your name as the sender, your note, the deck link and attached images. | United States (US East) |
| Anthropic | Claude models read the text you give us and arrange it into a deck, and describe photos for captions and alt text. | Pasted, résumé and job-description text, and photos as small 1024 px WebP copies (never your original files). | United States |
Added before each goes live
These are planned. Each one moves to the list above before it handles any of your data.
| Name | Purpose | Data it receives | Where |
|---|---|---|---|
| Google Gemini (paid tier) | Creates AI-styled versions of a photo when you ask for an AI style. | One re-encoded copy of the photo (at most 1600 px, metadata removed) and the style instructions. | Google Cloud (to be confirmed before it goes live) |
| OpenAI | Backup image generation when the first engine is unavailable, and automated moderation of text and images. | The same photo copy and instructions as above; text and images sent for moderation checks. | United States |
| AWS Rekognition | Moderation checks and face detection (face boxes, the one-face check, the likely-minor and celebrity flags described in the biometric notice). We will apply AWS's AI-services opt-out so this content is not used to improve AWS products. | A downscaled copy of each uploaded photo (at most 1600 px, metadata removed). | United States (AWS us-east-1) |
| fal | Cutouts: removes the background from a photo you choose. | A re-encoded copy of that one photo. We ask fal to expire its copy and keep only the result. | United States (to be confirmed before it goes live) |
| Thorn Safer or Microsoft PhotoDNA | Matches uploads against lists of known child sexual abuse material. We will use one of the two; which one is not decided yet. | A digital fingerprint (hash) of each uploaded image. | To be confirmed with the vendor |
Not subprocessors
The National Center for Missing & Exploited Children (NCMEC) is not a subprocessor. It is a recipient of legal reports: the law requires us to report apparent child sexual abuse material to it.
Questions
Write to [email protected]. The privacy policy says how long data is kept.