Login Get a Demo
01 Integration · OpenAI

Ship on OpenAI. Keep every prompt accountable.

OpenAI is already in your products, your code and your people’s browsers. Cranium ONE shows you where, inspects what’s being sent, and enforces your policy before a prompt becomes an incident.

02 Across the AI Trust Loop

What Cranium ONE does with OpenAI.

01Discover02Observe03Govern04Secure05Prove↻ Continuously
Discover

Find OpenAI everywhere it runs

CodeSensor records the OpenAI models your repositories call — in Python, TypeScript, Java, Go, C# and more, including the OpenAI .NET SDK and Azure OpenAI client. AgentSensor maps systems built on the OpenAI Agents SDK, down to their tools and MCP servers.

Observe

See who’s using it

Connect your SIEM and Shadow AI shows which AI services are in use across the company, who is using them and how much data is moving through them. Guardian adds the prompt-level view for the apps you build.

Govern

Policy per use case

A profile sets which signals matter for each use case — sensitive data, secrets, jailbreaks, prompt leaking, goal hijacking — and whether each one is blocked, modified or passed.

Secure

Stop it before it lands

In Enforce mode Guardian sits between your application and OpenAI, acting on every prompt and response in line. Listen mode lets you watch first, with no change to the request path.

Prove

Evidence on demand

Trace explains every verdict, AI Cards package the record for customers and auditors, and BOMs export to CycloneDX and SPDX.

03 Connect

Three steps. No rip-and-replace.

Cranium ONE works with the stack you already run. Your policy stays yours — we make it enforceable.

  1. 01
    Add the model

    In Guardian, add an OpenAI model with its project ID, API key and organization ID.

  2. 02
    Route and listen

    Send an application’s traffic through Guardian in Listen mode and watch real sessions arrive.

  3. 03
    Enforce with confidence

    Validate the profile with Cannon, then flip it to Enforce for in-line control.

04 Questions

Cranium ONE & OpenAI, answered.

What does Cranium ONE need to connect to OpenAI?

An OpenAI project ID, an API key and your organization ID. You add them when you create the model in Guardian, and the connection is scoped to that project.

Can Cranium show who in my company is using OpenAI and other AI services?

Yes. With your SIEM connected, Shadow AI shows which AI services are in use across your organization, who is using them and how much data is moving through them.

Does Cranium detect apps built with the OpenAI Agents SDK?

Yes. AgentSensor detects agentic frameworks in your repositories, including OpenAI Agents, and lists each agent with its model, instructions, tools, handoffs, guardrails and MCP servers in the AI Bill of Materials.

Will Guardian slow down or change my OpenAI calls?

Not in Listen mode, which evaluates a copy of the traffic in parallel. In Enforce mode Guardian runs in line so it can block, modify or pass prompts and responses according to your profile.