Where Salus sits
Two paths into one engine. Integrate Salus into the applications you build, or deploy it around the ones you cannot change.
Every protected AI interaction takes one of two shapes. Either there is a cooperating component in the request path, your application, your service, your AI gateway, that can hand the request to Salus deliberately. Or there is not, and Salus has to be inserted around the usage without the application’s cooperation.
Ownership only predicts the answer. You may own an application and still choose the endpoint path because changing it is expensive, or not own the application but control a gateway in front of it that can call Salus.
Integrated: Salus in your stack
When the answer is yes, your code or your gateway calls Salus locally. Salus runs inside your environment as a set of containers behind one gateway service; it is not a hosted service and it never becomes a cloud hop.
- Direct API. Your application calls
POST /v1/detect/textbefore the provider call andPOST /v1/recover/textafter it. Documents go throughPOST /v1/detect/file. The API reference has the shapes. - Your AI gateway. If you run LiteLLM, Salus plugs in as a Generic Guardrail on the pre-call and post-call hooks, streaming included. Your gateway keeps authentication, provider routing, credentials and model selection. See LiteLLM guardrail.
Identity is consistent across a workload. A context_id spans a conversation, a case or a batch run, and within it the same person is the same token in every call.
Endpoint: Salus on managed devices
When the answer is no, browser AI, desktop AI clients, specialist tools whose vendors will not add an integration, Salus is deployed on the device. What is typed or attached is protected before it leaves, and the answer is restored on the way back. The Endpoints chapter covers what IT and security teams should expect.
Side by side
| Integrated | Endpoint | |
|---|---|---|
| The situation | A component you control is in the request path | The application cannot be changed |
| Who does the work | Your application or gateway calls Salus | Salus protects on the device |
| Typical surfaces | In-house apps, services, analytics, agents, your gateway | Browser AI, desktop AI apps, specialist tools |
| Who adopts it | Engineering and AI platform teams | Security, IT and endpoint teams |
| Engine, tokens, contexts, rules | Shared | Shared |
What Salus does not decide
Salus does not decide which provider a request goes to, and it does not decide whether a request may leave at all. Your gateway decides where. Your DLP decides whether. Salus changes the form of what leaves so that what reaches the model is safe to send, and works alongside both.
All example values in this documentation are synthetic.