Skip to content
Docs

Security & Compliance

Interplay’s security model starts from where it runs: the platform is self-hosted, so flows, prompts, credentials, business data, and model traffic all stay inside the environment the organization controls. Everything below layers on top of that.

What Single Sign-On does Interplay support?

Section titled “What Single Sign-On does Interplay support?”

Interplay supports enterprise Single Sign-On using SAML. An administrator can configure the identity provider by hand or by uploading the IdP’s metadata XML directly, and the editor issues authenticated sessions through the provider. Local administrative accounts remain available as a recovery path if directory authentication is unavailable.

How is access controlled?

Section titled “How is access controlled?”

Access is role-based. Accounts are assigned roles that determine what they can see and do — administrators manage users, platform configuration, and the palette, while other users work within the flows and projects they are entitled to. Administrative API access is separately gated by API key or administrative credentials.

How are credentials and secrets protected?

Section titled “How are credentials and secrets protected?”

Interplay includes a secrets wallet: credentials and API keys are stored once in the platform and granted by role, rather than pasted into individual flows. Flows reference a stored credential instead of embedding it, which means:

  • Secrets are not exposed in exported flow JSON or in version control.
  • A credential can be rotated in one place without editing flows.
  • A flow author can use a connection without being able to read the secret behind it.

How is the runtime hardened?

Section titled “How is the runtime hardened?”
  • Non-root execution — the production container runs as an unprivileged user with the root account locked, a restrictive umask, and compatibility with OpenShift’s arbitrary-UID model.
  • Minimal image surface — the runtime image is built multi-stage and deliberately excludes utilities with known vulnerability history, with hardened interpreter settings applied.
  • User data protection — flows, configuration, authentication settings, and installed nodes are explicitly preserved across image and package upgrades.
  • Ongoing hardening — the platform is subject to regular dependency and vulnerability review, with security fixes shipped as part of the normal release stream.

What about data residency and regulated environments?

Section titled “What about data residency and regulated environments?”

Because Interplay is deployed on infrastructure the customer owns — including air-gapped sites with no outbound connectivity — data residency is a property of where the deployment sits. There is no vendor-side telemetry requirement, no external control plane, and no obligation to route model traffic through a third party: model calls can be pointed at a private inference engine such as Lifeboat so prompts and completions never leave the boundary either.

How do you govern and audit AI usage across the enterprise?

Section titled “How do you govern and audit AI usage across the enterprise?”

Interplay secures the build-and-run layer. For enterprise-wide AI governance — per-team budgets, data-loss prevention on prompts and responses, guardrails, model-usage audit trails, and shadow-AI discovery across every provider and application — pair Interplay with AgentWatch, which sits in front of model traffic as a governance gateway.

Sources and references

Section titled “Sources and references”