Glossary

What is an LLM gateway?

An LLM gateway is a routing layer that accepts AI requests from an application and forwards them to one or more model providers.

Definition

An LLM gateway is a routing layer that accepts AI requests from an application and forwards them to one or more model providers.

Definition

LLM gateway — short definition

Why it matters

Why this matters

Without a gateway, every client calls providers directly. There is no central place to mask sensitive data, apply policy, rotate keys or observe usage. A gateway gives security and platform teams one enforcement point for all AI traffic.

Without a gateway, every client calls providers directly. There is no central place to mask sensitive data, apply policy, rotate keys or observe usage. A gateway gives security and platform teams one enforcement point for all AI traffic.

How it works

How it works

The application sends a raw prompt to the gateway. The gateway replaces sensitive values with placeholders and forwards the masked prompt to the LLM provider. The provider returns a response with placeholders. The gateway rehydrates placeholders to the original values before returning the response to the application. The provider never sees original values.ApplicationRaw promptPrivian gatewayMask · Route · RehydrateLLM providerSees masked prompt onlypromptmasked promptresponse w/ placeholdersrehydratedBYOK trust boundary
Prompt path through a privacy-first gatewayOriginal values never cross the BYOK boundary.

Framework

LLM gateway process

  1. 01

    Accept

    The gateway accepts HTTPS/JSON requests from your app (prompt + model).

  2. 02

    Resolve

    It picks the provider and model based on the request and your configuration.

  3. 03

    Forward

    It calls the provider with your credentials and returns the response.

Implementation

Learn how this works in Privian

From definition to implementation, docs and architecture — the same idea at different layers.

FAQ

Frequently asked questions

What is an LLM gateway used for?
Centralizing routing, authentication, model selection, observability and — in a privacy-first gateway — masking and prompt security for all AI traffic.
Is an LLM gateway the same as an AI gateway?
The terms overlap. 'AI gateway' is often used more broadly; 'LLM gateway' specifically denotes a routing layer for large-language-model traffic.
Do I need an LLM gateway if I only use one provider?
Yes if you want a single enforcement point for masking, prompt security, observability and key rotation. No if you only need basic routing.

Where to go next

From definition to implementation

This entry is part of the Privian content library. If you are evaluating how this concept is implemented in practice, the links below are the most direct paths.

Related product

Privian LLM Gateway

A privacy-first LLM gateway with PII masking, BYOK and rehydration applied at the edge.

Open product page →