Skip to content
NewFinOps, Kubernetes and AI, on one platform
Book a demo
UnitVane docs
API referenceReleases

Connect AI providers

Read access to AI cost and token telemetry on Azure, AWS and Google Cloud, plus OpenAI and Anthropic organizations.

On this page

Azure OpenAI and Azure AI Foundry

UnitVane AI reads amortized cost from Cost Management and token and request metrics from Azure Monitor for every Azure OpenAI deployment. Assigning roles at the root management group lets it discover every subscription with AI spend automatically; list sandbox subscriptions to exclude them.

Azure CLI
# Read Azure OpenAI cost and token metrics across the tenant.# Assign at the root management group so subscriptions with AI spend are discovered automatically.MG=$(az account management-group list --query "[?displayName=='Tenant Root Group'].name" -o tsv)az role assignment create --assignee "<kubeon-ai-identity-client-id>" \  --role "Cost Management Reader" --scope "/providers/Microsoft.Management/managementGroups/$MG"az role assignment create --assignee "<kubeon-ai-identity-client-id>" \  --role "Monitoring Reader" --scope "/providers/Microsoft.Management/managementGroups/$MG"az role assignment create --assignee "<kubeon-ai-identity-client-id>" \  --role "Reader" --scope "/providers/Microsoft.Management/managementGroups/$MG"

Amazon Bedrock and SageMaker

Cost comes from the same CUR 2.0 export KubeOn uses. Add this policy to the KubeOn billing role so UnitVane AI can read Bedrock invocation and token metrics and list provisioned throughput.

IAM policy
{  "Version": "2012-10-17",  "Statement": [    {      "Sid": "BedrockUsageMetrics",      "Effect": "Allow",      "Action": ["cloudwatch:GetMetricData", "cloudwatch:ListMetrics"],      "Resource": "*"    },    {      "Sid": "BedrockInventory",      "Effect": "Allow",      "Action": [        "bedrock:ListProvisionedModelThroughputs",        "bedrock:ListInferenceProfiles",        "bedrock:ListFoundationModels",        "sagemaker:ListEndpoints"      ],      "Resource": "*"    }  ]}

Google Vertex AI

Cost comes from your BigQuery billing export. Grant the service account read access to Cloud Monitoring and Vertex AI inventory.

gcloud
# Vertex AI cost comes from the same BigQuery billing export as the rest of the bill.gcloud projects add-iam-policy-binding acme-prod \  --member="serviceAccount:kubeon-ai@acme-billing.iam.gserviceaccount.com" \  --role="roles/monitoring.viewer"gcloud projects add-iam-policy-binding acme-prod \  --member="serviceAccount:kubeon-ai@acme-billing.iam.gserviceaccount.com" \  --role="roles/aiplatform.viewer"

OpenAI and Anthropic APIs

Add an admin or usage-read API key for each organization in AI > Providers. UnitVane AI reads daily cost and usage by project and model. Traffic through the gateway is metered per request in addition.

Owners and tags

DimensionDefault tag keys
Teamteam, Team, owner_team
Business ownerowner_business, BusinessOwner
Technical ownerowner_technical, TechnicalOwner
Environmentenvironment, env
Product, zone, tenantMapped from tags or gateway keys in Settings > Dimensions
Need help with this guide?Contact support