Privacy and data handling
What ORYNX stores, what it doesn't, and who else processes your requests.
Request content
Prompts, completions, files, images, tool calls and all other request and response content pass through ORYNX in memory. ORYNX doesn't write them to a database, to disk or to logs.
What ORYNX keeps
For billing and support, ORYNX records metadata about each request:
- time, model, API format and whether it streamed
- token counts (input, cached, cache writes, output, reasoning) and the cost
- duration, status and an error category if it failed
- the kind of client, from the
User-Agentheader (for example Claude Code or curl) - the calling IP address shortened to its network (
/24for IPv4,/48for IPv6) - the ORYNX request id and the id the provider returned, used to reconcile billing
When a model provider returns an error, ORYNX keeps that error message for up to 7 days to investigate the problem. Some providers quote part of the request in their error messages. Hourly usage totals feed the charts in Key lookup. The privacy policy says how long each kind of data is kept.
Model providers
To answer your request, ORYNX forwards it to the provider that runs the model. Providers process your content under their own terms and may keep it for a period, for example for abuse monitoring. Don't send data you are not allowed to share with a third-party AI provider.