When an ISO 27001 auditor asks for proof that every AI‑driven operation is controlled, the organization must demonstrate who invoked the OpenAI Agents SDK, what prompts were sent, what responses were returned, and whether any sensitive data was masked.
Current practice without a gateway
Most teams embed the OpenAI Agents SDK directly into their services and authenticate with a static API key or a service‑account token. The key lives in environment variables, configuration files, or secret managers that the application reads at runtime. Because the SDK talks straight to OpenAI’s endpoint, the organization loses visibility into:
- Which user or automated process triggered a request.
- What exact prompt and response were exchanged.
- Whether any sensitive output left the system unredacted.
- Any post‑hoc evidence that the request complied with internal policies.
This "direct‑connect" model satisfies authentication but provides no audit, no just‑in‑time approval, and no inline data protection. ISO 27001 clauses that require controlled access and auditable activity remain unmet.
What the precondition solves – and what it leaves open
Introducing a non‑human identity (a service account) that has the minimum permissions needed to call the SDK is a necessary first step. The service account is federated through OIDC, so the platform can prove who the request originates from. However, the request still travels straight to OpenAI, bypassing any enforcement point. The organization still lacks:
- Real‑time inspection of the request payload.
- Automatic masking of personally identifiable information in responses.
- Recorded sessions that can be replayed for forensic analysis.
- Human approval for high‑risk prompts.
In ISO 27001 terms, the control environment is still incomplete: access is identified, but the evidence of how that access was used is missing.
How hoop.dev generates evidence for iso 27001
hoop.dev acts as a Layer 7 identity‑aware proxy that sits between the OpenAI Agents SDK and the OpenAI service. The gateway is the only place where traffic can be inspected, approved, masked, and recorded. Because the gateway holds the credential, the SDK never sees the raw API key.
When a request passes through hoop.dev, the gateway produces the following enforcement outcomes:
- Session recording: hoop.dev logs every request and response with timestamp, user identity, and request metadata. The logs are immutable and can be exported for audit, providing a tamper‑evident audit trail.
- Inline data masking: hoop.dev redacts sensitive fields identified by policy (for example, credit‑card numbers or SSNs) before the response leaves the gateway, ensuring downstream systems never store raw PII.
- Just‑in‑time approval: hoop.dev triggers a workflow that requires a designated approver to consent before high‑risk prompts proceed.
- Command blocking: hoop.dev rejects disallowed operations (such as prompts that attempt to extract model weights) at the gateway, preventing misuse.
Each outcome maps directly to ISO 27001 control objectives:
- A.12.4 Logging and monitoring – hoop.dev’s immutable session logs provide the required evidence of user activity.
- A.9.2 User access management – the gateway enforces least‑privilege access and records who exercised that access.
- A.18.1 Protection of records – masked data ensures that records containing personal data are protected in accordance with the standard.
- A.6.1.2 Segregation of duties – just‑in‑time approvals introduce a second party to validate high‑risk actions.
Because hoop.dev lives in the data path, the organization can demonstrate to auditors that every AI‑driven operation is both controlled and traceable, satisfying the evidence requirements of ISO 27001 without additional tooling.
Getting started with hoop.dev for the OpenAI Agents SDK
The onboarding flow follows the standard hoop.dev quick‑start. Deploy the gateway (Docker Compose or Kubernetes), configure the OpenAI connection, and point the SDK’s endpoint to the gateway URL. Authentication uses OIDC, so the same identity provider that powers your internal SSO can be reused.
For detailed steps, see the getting‑started guide and the broader learn section that explains policy authoring, masking rules, and approval workflows.
FAQ
Does hoop.dev replace the OpenAI API key?
Yes. The gateway stores the credential and presents it to OpenAI on behalf of the SDK. The SDK never receives the raw key, eliminating accidental exposure.
Can I retain the original logs after a breach investigation?
hoop.dev’s session logs are immutable by design and can be exported to any long‑term storage solution you choose, providing a reliable evidence chain for post‑incident analysis.
Is hoop.dev itself ISO 27001 certified?
hoop.dev does not claim certification. It generates the audit‑ready evidence that helps your organization meet ISO 27001 requirements when you run the OpenAI Agents SDK through it.
Explore the source code and contribute to the project on GitHub.
