Sub-processors
Draft — have a lawyer review before you take payment.
Model inference is per-call and per-conversation: evidence goes only to the provider whose model was selected for that call. A provider your workspace never selects never receives anything.
| Processor | Purpose | Data |
|---|---|---|
| Anthropic | Model inference, when a Claude model is selected | Evidence in the prompt for that call |
| OpenAI | Model inference, when a GPT model is selected | Evidence in the prompt for that call |
| Model inference, when a Gemini model is selected | Evidence in the prompt for that call | |
| OpenRouter | Model routing, only if we enable it | Evidence in the prompt for that call |
| PayPal | Payments | Workspace id, workspace name, plan and amount of each payment. Payment details are entered at PayPal and never reach our servers |
| SendGrid (Twilio) | Transactional email: verification, password reset and invitations | Recipient email address and the message, including its link |
| OVHcloud (France) | The dedicated server that runs the application and its database | All service data, stored on our server in its data centre |
Hosting
The application and its database run on a dedicated server we administer ourselves, rented from OVHcloud in France. There is no third-party analytics, error-reporting or session-recording service on this site.
Changes
We will give 30 days’ notice by email before adding a sub-processor that processes evidence from your machines.