Change one line, protect every prompt.
The Redact proxy anonymises personal data before it reaches any model, then restores it in the response. You point your client at a new base URL, and sensitive fields never leave in the clear.
Redact sits in front of your model provider. Change the base URL your client uses and the proxy handles anonymisation and de-anonymisation transparently.
On the way out, personal data is replaced with placeholders. On the way back, the model's response is de-anonymised, so your application still sees real values.
Because it works at the API boundary, the proxy protects prompts regardless of which model you send them to.
Related capabilities on the same authorisation layer. Every one is enforced before an action runs and sealed to the Provenance Vault.
Put one workflow behind Xybern and watch every agent action authorised, and sealed to the vault.
Xybern