Sub-processor List

BeMyWords — operated by Skiwo AS Effective date: 2026-09-20 Last updated: 2026-09-20

This page lists the sub-processors Skiwo AS engages to operate the BeMyWords service. Sub-processors are third parties we rely on to deliver the Service, each bound by data protection obligations equivalent to those in our Data Processing Agreement. Suppliers Skiwo AS uses only for its own internal operations, such as accounting software, do not process your data and are not listed here.

We will notify customers of material changes to this list at least 30 days in advance via email to the workspace owner. Customers may object to a proposed change within that period; if the objection cannot be resolved, the customer may terminate their subscription.

Infrastructure

Sub-processor	Purpose	Data processed	Location
Heroku (Salesforce)	Application hosting, managed Postgres, managed Redis, file attachments, automated database backups	All workspace content, account data, operational data, uploaded screenshots, system error logs	EEA

Heroku is the only cloud-infrastructure sub-processor Skiwo AS engages directly. Heroku in turn runs on AWS infrastructure under Heroku's commercial terms; that is Heroku's choice of underlying provider, not a separate engagement by Skiwo.

Email delivery

Sub-processor	Purpose	Data processed	Location
Amazon Simple Email Service (SES)	Transactional email — account notifications, invitations, password resets	Email addresses, email content	EEA (eu-central-1)

Machine translation and vision

Sub-processor	Purpose	Data processed	Location	Safeguards
Anthropic (via Heroku Inference)	Machine translation via Claude; vision-based string matching when a customer uploads a screenshot for in-context translation	Source strings submitted for translation; uploaded screenshot images and the project's source-language strings, when the screenshot feature is used; context provided by customer	Routed via Heroku (EEA)	Anthropic API terms (no training on submitted data); Heroku acts as delivery path

BeMyWords does not use OpenAI, DeepL, or any other external MT providers. Claude is accessed through Heroku's inference routing; customer data is processed by Anthropic under their commercial API terms, which exclude the data from model training.

The vision model reads screenshots by fetching them over the network rather than receiving them inline, so an uploaded image is briefly readable at a URL. Skiwo holds the image in its Redis cache under a single-use, unguessable token, the model fetches it once, and the entry is deleted as soon as the model responds — or expires by itself within three minutes if it is not.

Only the strings the model extracts, and the screenshot's own metadata, are written to the application database. The image is not stored there, and no copy is kept once the request has finished. The uploader's browser also keeps a local copy in IndexedDB for up to 8 hours so the image remains visible during a single session; that local cache is the user's browser, not a sub-processor.

Billing and invoicing

Sub-processor	Purpose	Data processed	Location
Payout Partner AS (org. no. 922 810 699, Pilestredet 17, 0164 Oslo)	Invoice issue, delivery and payment records for paid plans	Billing name and email, billing address, organisation or VAT number, and invoice records. The Service takes no card payments, so no card data is processed.	Norway (EEA)

Payout Partner AS processes billing data only for customers on a paid plan. Customers on the unlimited internal plan, and customers we invoice by hand, do not have billing data processed by Payout Partner AS.

Error tracking and observability

Error tracking is handled internally — exception details are persisted to the application database (the system_errors table) on Skiwo's own infrastructure. No third-party error-tracking service (e.g. Sentry, Bugsnag, Rollbar) is engaged, so customer data in stack traces stays within Skiwo's systems.