SCAI Enterprise API

Swiss Cheese AI logoby Swiss Cheese AI

Model Council as a Service for teams that need independent AI generation, challenge and synthesis without moving confidential work into another SaaS data store.

Customer-hosted runtime

Deploy SCAI inside your VPC, Kubernetes cluster or private network. The runtime is configured by the customer and does not persist prompt or document content.

01

Your security boundary

The Model Council runs where you choose. Customer data does not need to enter Swiss Cheese AI infrastructure.

02

Approved models only

Model destinations are deployment configuration. API callers cannot replace or redirect the configured model endpoints.

03

No content persistence

Requests are processed in memory. Runtime logs can remain limited to operational metadata rather than customer content.

How the private council works

The orchestration stays consistent while the customer controls deployment, network policy and model access.

Private runtime
Your productTraceable, internal app or workflow
→
SCAI RuntimeGenerator → evaluators → integrator
→
Approved modelsPrivate, local or enterprise endpoints
Policy boundary: customers can enforce egress with their own firewall, Kubernetes NetworkPolicy or service mesh. Model URLs are not accepted from API request payloads.

One endpoint. Multiple independent opinions.

Your application sends one request to the local SCAI runtime and receives an integrated council result.

OpenAPI 3.1
POST/v1/council/run

Request

Authorization: Bearer <runtime-token>
Content-Type: application/json

{
  "input": "Review this verification protocol for unsupported claims.",
  "context": "Optional customer context",
  "metadata": {
    "document_id": "VP-104"
  }
}

Response

{
  "output": "Integrated council result...",
  "council": {
    "generator": { "model": "..." },
    "evaluators": [
      { "model": "..." },
      { "model": "..." }
    ],
    "integrator": { "model": "..." }
  },
  "policy": {
    "persisted": false,
    "requestDefinedModels": false
  }
}
OpenAPI specificationService health

Deployment model

Choose how closely the runtime needs to stay inside the customer's architecture.

ModeWhere SCAI runsCustomer content reaches SCAI cloud
✓Cloud APISwiss Cheese AI infrastructureYes
✓Private RuntimeCustomer VPC / cluster / private networkNo
✓Air-gappedCustomer-controlled isolated environmentNo

Security notes

Does Swiss Cheese AI receive private-runtime prompts?

Not by design. The private runtime processes the request inside the customer's environment and can be restricted by customer-controlled network policy so prompt and document content never routes to Swiss Cheese AI infrastructure.

Can an API caller choose an arbitrary model endpoint?

No. Model endpoints are deployment configuration, not request parameters. This prevents application callers from redirecting sensitive content to an unapproved destination.

What remains under customer control?

Network egress, model credentials, approved endpoints, infrastructure logging, encryption, retention and audit storage can remain under the customer's own controls and policies.

Model Council as a Service, without giving up your security boundary.

Swiss Cheese AI•Enterprise API•OpenAPI