Subprocessors
Last updated 2 August 2026. Lumino Solution LLC, trading as Cala. Version history.
This is the complete list of third parties that process data on behalf of Lumino Solution LLC, trading as Cala. It is kept in version control and published here. Anyone assessing Cala can read the whole list without having to ask us for it. That includes a colleague, an IT team or a procurement reviewer.
1What a subprocessor is
A third party that processes data on our behalf, and why the list is public.
A subprocessor is a third party we use to process personal data or customer content on our behalf. That is the sense given in Article 28 of the General Data Protection Regulation. Each one below is under a written data processing agreement, is used for a stated purpose, and receives only what that purpose requires.
The register is deliberately short. Every entry is one more place your data can be, so an entry has to earn itself. There is no font content delivery network, no session recorder, no advertising pixel and no third-party embed on this site. The one analytics entry runs cookieless with consent denied by default, which is why the site still shows no consent banner.
2Current subprocessors
Four, with what each one receives and where it is processed.
| Subprocessor | Purpose | Data processed | Region |
|---|---|---|---|
| Anthropic, PBCModel inference | Chat responses and document extraction. | Chat message content, and the text and images of the specific document window being processed. Both transit our stateless proxy in memory and are not stored by us. | United States |
| SupabaseAuth, Postgres, Edge Functions | Account identity, account and billing metadata, and the server-side proxy functions. | Email and auth identity, subscription tier and status, usage metering counters, credit balance. No chat content and no document content. | United States |
| StripePayments | Subscription billing and metered overage charging. | Name and email, payment method held by Stripe, subscription and invoice records. | United States |
| Google LLCWeb analytics (cookieless) | Aggregate traffic measurement on calalayer.com: which pages are visited, and from where. | Consent-denied, cookieless pings only: page URL, referrer, and coarse device and region signals. No cookies are set, no advertising features are enabled, and nothing ties a visit to an account. | United States |
| VercelWebsite hosting | Serving calalayer.com. | Request metadata: IP address, user agent, requested URL and timestamp, as part of ordinary web serving. | United States |
International transfers
All four subprocessors process data in the United States. Some transfers carry personal data of a person in the European Economic Area or the United Kingdom. Those transfers are made under the Standard Contractual Clauses incorporated in that subprocessor’s data processing agreement. The transfer basis is set out in the privacy policy.
Providers that are not in use
Adapters for other model providers exist in the codebase, but Cala sends all inference to Anthropic today. No other provider processes anything. If that changes, the provider is added to this register and notified before it starts processing, on the terms in section four below.
3What never reaches a subprocessor
Your documents, your knowledge base, your chat history and your investigations.
Cala runs on your computer. Your documents, and everything Cala learns from them, stay there. Extracted knowledge, chat history and investigation content live in a SQLite database on your own machine. We do not upload them and we do not store them on our servers, and none of it is handed to any subprocessor as a body of data. That is the reason this list is as short as it is.
Here is the part that is easy to leave out. To answer a question, that message and the specific document pages being read are sent to the inference provider through Cala’s proxy, over TLS. They pass through in memory and are not stored on our side. Anyone telling you that nothing at all leaves your machine is describing a product that cannot answer a question.
Cala stores no document content and no chat content on its servers. That is also why there is nothing of yours on our servers to train a model on, and Cala has no model of its own to train.
4How changes are notified
You hear about a new subprocessor before it starts processing, not after.
When this register changes, the change is made in version control first and published here with a dated entry in the legal changelog.
- New subprocessors are notified in advance. Registered account holders are told by email before a new subprocessor begins processing any data, not after it has started.
- Removals and scope changes are recorded too. If a subprocessor is dropped, or the categories of data it receives change, the register and the changelog are updated in the same change.
- The dated version above is the current one. There is no separate internal list.
5Questions and objections
Who to write to, and what happens if you object.
If you object to a proposed new subprocessor, write in before it takes effect. If the objection cannot be resolved, you can cancel your subscription from your account page and it stops at the end of the period you have paid for. Your documents, your knowledge base, your investigations and your reports stay on your machine, because we never held a copy of them.
Related pages: privacy policy, how Cala protects your data, intended use and AI disclosure.