Integration guide / VISCOR

Purpose-bound fields for AI workloads

VISCOR runs inference and manages quotas and its usage ledger. NSECURED MemberGo manages protected fields and access policy. The two systems deploy independently and interoperate through authenticated interfaces.

Responsibility model

Where compute meets data control

Compute platform

VISCOR

Retains field tokens and purpose, and binds each deployment to a service account. The workload requests data when it needs a value.

Authenticated requests & responsesmTLS

Data protection platform

NSECURED MemberGo

Stores encrypted fields, checks the tenant, purpose and policy, and records the access before Key Broker–controlled release. Successful accesses are exported as usage per tenant.

Operational interfaces use mutual TLS and operations credentials. Field access uses the service identity bound to the tenant and deployment.

Reference flow

A quality audit from setup to reconciliation

01

Provision the tenant and purpose

Register the quality-audit purpose and process-parameter type. Set the model, time and access limits and grant the deployment service account the necessary scope.

02

Submit field references

The job sends field tokens, its workload reference, model version and purpose. The tokens link the records; the authenticated service still needs an applicable grant.

03

Enforce the release decision

If all conditions pass, approved values are used only in memory for that run. An unauthorized purpose or field stops the job and returns a data-access denial.

04

Synchronize changes and reconcile usage

Signed events communicate tenant changes. A cursor feed catches missed events. Successful accesses are pulled separately per tenant; each counts as one vault request.

Design review

Three decisions for an integration review

01

Which deployment may read?

Identify the workload, allowed model and service account. Name a business owner for the permitted purpose.

02

Which field will prove the policy?

Select an eligible custom type and use synthetic data to exercise both allowed and denied decisions.

03

Which outputs need review?

Review the job output, logs and ledger. Confirm raw values are not persisted and only successful requests count toward usage.

Bring the integration boundary into your PoC

Use synthetic fields to verify an allowed request, a whole-request denial and the associated audit evidence.