Skip to main content

​
Cloud AI Providers

Cloud AI Provider connectors let SecureAI reach into the AI platforms your organization already uses — OpenAI Platform, Anthropic Console, Azure AI Foundry, Google Vertex AI, and AWS Bedrock — and automatically inventory everything running there. No agent or SDK needs to be deployed on the provider side; SecureAI polls each provider’s management API with read-mostly credentials you supply. Once connected, a connector discovers and continuously syncs:
  • Agents, models, and deployments running on the platform → shown in Cloud Sensors.
  • Non-Human Identities (NHIs) — API keys, service accounts, and (where supported) BYOK keys → shown in NHI Inventory, with block/revoke controls.
  • Usage, tokens, and cost for the last 30 days.
  • Governance signals — IAM/RBAC bindings, spend limits, and audit logs, surfaced in per-provider Insights.

​
The setup loop

Every provider follows the same four steps:
  1. Open Admin → Integrations and select the Cloud category.
  2. Connect the provider card and enter its credentials (see each provider’s page).
  3. Test the connection to validate the credentials.
  4. Sync — the first sync runs a full import; subsequent syncs are incremental and also run on a schedule.
Where things appearConnector setup lives in Admin → Integrations. The inventory it produces appears in Admin → Agent Registry: discovered agents/models under Cloud Sensors, and discovered identities under NHI Inventory. You can re-run a sync from either place.

​
Choosing credentials

Use read-only / viewer credentials wherever possible — SecureAI only needs to read your inventory to build the picture. Some optional capabilities (revoking a leaked key, creating a spend alert) require additional write permissions; each provider page calls these out explicitly, and they are always optional. All secrets you enter (client secrets, service-account JSON, admin API keys) are encrypted at rest.

​
Provider guides

​
Inference on your own account

Besides inventorying, SecureAI can run LLM traffic on your own Google Cloud account, so Google bills you directly and you can use your committed spend (CUDs). It is a separate integration, under Integrations → AI providers: Google Vertex AI — Inference on your account. The same applies to Azure: Azure AI Foundry — Inference on your account runs LLM traffic on your own Foundry resource so Microsoft bills you and you draw down your Azure commitment (MACC). And on AWS, Amazon Bedrock — Inference on your account runs LLM traffic, Claude included, on your own AWS account so AWS bills you and you draw down your AWS commitment (EDP).