Skip to content
VeraFrame Docs

Integration — overview

Integration is VeraFrame’s tier for organizations that cannot or do not want to run on shared SaaS infrastructure. Instead of using the hosted SaaS Cloud, Integration deploys VeraFrame into your own AWS or Azure tenancy. Your data, your models, your logs — all stay in your environment. VeraFrame supplies the verification engine; you supply the cloud.

Integration is the right tier when:

  • Your data cannot leave your cloud for regulatory, contractual, or internal-policy reasons.
  • You want VeraFrame’s verification pipeline as a building block inside a larger AI platform you own.
  • You have existing MLOps, observability, and security tooling that should also cover the verification layer.
  • You are subject to EU AI Act high-risk requirements and need full control over the compliance boundary.

If none of those apply, the SaaS Cloud tier is usually a faster and simpler fit.

What is deployed

The Integration package includes:

  • Verification engine — the deterministic validator, retrieval-and-scoring pipeline, document-template renderer.
  • REST API — programmatic access to all four modes (Generate, Audit, Ask, JSON API) and to validation history, templates, source groups, and audit events.
  • Admin and user UI — the same web interface SaaS Cloud tenants use, running in your tenancy.
  • Webhook dispatcher — outbound notifications for validation events.
  • All compliance controls — the same feature-flag set as Compliance Edition, configurable to your needs.

Integration points

VeraFrame is designed to fit into an existing enterprise environment rather than replace parts of it. Typical integration points:

  • Identity provider — OIDC / SAML integration for SSO.
  • Source systems — S3 (native) or REST APIs (via API-backed source groups).
  • Downstream consumers — the REST API and webhooks push validated output into CRM, ERP, ticketing, or AI assistant systems.
  • Observability — Cloud-native logs and metrics via CloudWatch or Azure Monitor; optional integration with your own SIEM.
  • Governance — external review workflow hands validations off to your legal or compliance platform.

In enterprise JWT/OIDC deployments, VeraFrame can also use your existing roles or claims to limit which source groups different users are allowed to query. This keeps source access aligned with your current identity model instead of introducing a separate data-permission system inside VeraFrame.

Deployment target

  • AWS — Lambda + API Gateway + DynamoDB + S3 + Bedrock. Deployed via infrastructure-as-code into your AWS account.
  • Azure — Equivalent components on Azure (Functions + Cosmos DB + Blob Storage + Azure OpenAI, with the same verification model).

Both deployments run in an EU region by default.

What this docs section covers

The pages in this Integration section are technical references for integrators — backend engineers, platform teams, and the people who will operate VeraFrame in your tenancy.

  • REST API — endpoint reference, authentication, mode parameters.
  • Webhooks — event types, payload shape, configuration.
  • Deployment — how VeraFrame is installed into your cloud.

For the concepts and product behavior, the main docs sections still apply — Integration customers use the same modes, the same trust report, and the same compliance controls as SaaS customers. Integration just changes where VeraFrame runs.

One important enterprise-specific difference is source access: Integration deployments can enforce source-group restrictions and optionally forward selected user context to downstream REST APIs when your architecture needs that.

Getting started with Integration

Integration is a sales-engaged process, not a self-serve signup. Contact michael@veraframe.io to scope a deployment. Typical first steps:

  1. Scoping call — your use case, data volumes, compliance profile.
  2. Architecture review — how VeraFrame fits in your environment.
  3. Proof-of-concept deployment — a narrow deployment on a test subset to validate integration.
  4. Production rollout — full deployment, data onboarding, user onboarding.