KoltrixSubprocessors
Everyone we rely on who could touch your data, what they do, and what actually reaches them.
Last updated
We give at least 30 days' notice by email before adding or replacing a subprocessor, as set out in our Data Processing Addendum. To be told when this list changes, email [email protected] and ask to be added to the notification list.
Currently engaged
| Provider | What it does | What reaches it | Where |
|---|---|---|---|
| OVHcloud | Servers, network and storage for the entire service | All customer data: mail, attachments, account records | Frankfurt, Germany (European Union) |
| Cloudflare | DNS, TLS termination and protection against denial-of-service attacks | Request metadata: IP addresses, URLs, headers. Not mail content at rest | Global edge network |
| OpenRouter (TypeSafe's Jev model) | Spam scoring and message classification | Message headers, subject and a truncated body. No attachments. Only when AI is enabled for the workspace | Provider-operated infrastructure |
| OpenRouter (Google Gemini and DeepSeek models) | AI summaries and draft replies | Message headers, subject and a truncated body with quoted history stripped. No attachments. Only when AI is enabled and the feature is used | Provider-operated infrastructure |
| Dodo Payments | Merchant of record: payment processing, invoicing, sales tax and GST | Billing name, email, country, card details (handled by them, never by us) | Provider-operated infrastructure |
A note about the AI providers
Two of the entries above receive the content of your email. We would rather state that here than leave you to infer it:
- Classification and spam scoring send headers, subject and a truncated body. This happens for every inbound message while AI is enabled.
- Summaries and drafts send the same, with quoted history stripped, when you or an agent asks for one.
- Attachments are never sent to either.
- Neither is permitted to train on your content under our agreement with them.
- Turning AI off in workspace settings stops both entirely. Sorting then uses our own spam engine, running on our own servers.
Named in our plans, not yet processing data
These appear in our architecture documents but are not yet engaged, so no customer data reaches them today. They are listed for transparency, and they will move into the table above — with notice — before that changes.
| Provider | Intended purpose | Status |
|---|---|---|
| an off-site object store (Backblaze B2 or Cloudflare R2) | Off-site encrypted backups | Not engaged |
What we do not use
- No advertising networks, and no data brokers.
- No third-party email infrastructure — Koltrix runs its own mail servers rather than reselling someone else's.
- No third-party session recording, heatmaps or behavioural tracking inside the product.