Skip to main content
Inviolet integrates with the OpenAI surface on three layers: the OpenAI SDK (server-side agents calling /responses or /chat/completions), the Assistants API (function-calling tools), and ChatGPT Enterprise custom actions invoked from the chat UI.

Why it matters

OpenAI tool calls share the same intent-extraction need as Claude. Wrapping the SDK at the call site yields the same decision feed entry shape — so dashboards, audit exports, and policies work identically across LLM providers.

1. Wrap the OpenAI SDK

This uses the evaluate path (intent.evaluate) — a lightweight pre-call check. For structural, credential-brokered enforcement where the agent never holds the real credential, mint a mandate and guard the call with @inviolet/agent-sdk-core — see the Quickstart.

2. ChatGPT Enterprise custom actions

ChatGPT Enterprise actions are OpenAPI-described HTTP calls. Point the action’s base URL at the Inviolet gateway path — Inviolet evaluates intent, then proxies the request to the underlying API:
Every action invocation now flows through Inviolet’s intent extractor before reaching your-app.

3. Verify

Open the decision feed. Trigger an action from ChatGPT Enterprise (or a tool call from the SDK). The event arrives with intent_class and the matched intention. Provider doesn’t matter — Inviolet’s view is unified.