How it works

One gateway between your app and the model

CloakGate is a control point in front of AI models. You keep your request shape, choose a known model, and sensitive data is handled before anything leaves for the provider.

Your app
one API key
CloakGate
detect · cloak · audit
GPT · Claude · Gemini
cloaked request
response returned to your app · metadata logged, prompt content not stored by default
01
Your app sends a request to CloakGate

Use one CloakGate API key and your normal request shape. No provider account required.

02
CloakGate detects sensitive data

Names, emails, phone numbers, invoices, secrets and more, based on the policies you enable.

03
Your rules are applied

Each detected value is redacted, pseudonymized or the request is blocked, per your policy.

04
The cleaned request is sent to the selected model

GPT, Claude, Gemini or another supported model receives only the cloaked request.

05
The response is returned

The model response is passed back to your application.

06
Usage and metadata are logged

Project, model, tokens, detected types and policy are stored. Prompt content is not stored by default.

See cloaking on a real prompt

request.prompt

Send a reply to Customer_1 at Email_1 about invoice Invoice_1. Card on file [BLOCKED_SECRET].

● pseudonymized● redacted● blocked→ sent to model

Cloaking is not magic. It is a configurable risk-reduction layer that helps prevent unnecessary sensitive data exposure. It does not guarantee that every sensitive value is removed in every case.

Start using AI through a safer gateway

Prepaid credits, known models, sensitive data cloaked before it leaves your app. Join the developer preview.