For the complete documentation index, see llms.txt. Markdown versions of all docs pages are available by appending .md to any docs URL.

Run agentgateway on Azure and reach Azure OpenAI with a managed identity instead of an API key.

Run agentgateway on Azure Container Apps or AKS, and reach Azure OpenAI with the managed identity that Azure already attaches to the workload. No API key goes into your configuration file.

Authenticate with a managed identity

Azure supplies credentials to the container through a managed identity on Container Apps and through workload identity on AKS. Agentgateway obtains an Entra ID token from whichever one is present.

# yaml-language-server: $schema=https://agentgateway.dev/schema/config

llm:
  models:
  - name: "*"
    provider: azure
    params:
      model: gpt-4o
      azureResourceName: my-resource
      azureResourceType: openAI
    auth:
      azure:
        implicit: {}

# yaml-language-server: $schema=https://agentgateway.dev/schema/config

llm:
  models:
  - name: "*"
    provider: azure
    params:
      model: gpt-4o
      azureResourceName: my-resource
      azureResourceType: openAI
    auth:
      azure:
        explicitConfig:
          managedIdentity:
            userAssignedIdentity:
              clientId: 00000000-0000-0000-0000-000000000000

Review the following table to understand this configuration.

For client secret and workload identity configurations, and for the Azure AI Foundry endpoint type, see Azure.

Run on Azure Container Apps

Run agentgateway as a serverless container with a user-assigned identity.

az containerapp create \
  --name agentgateway \
  --resource-group my-rg \
  --environment my-env \
  --image cr.agentgateway.dev/agentgateway:latest-dev \
  --target-port 4000 \
  --ingress internal \
  --user-assigned /subscriptions/<sub-id>/resourceGroups/my-rg/providers/Microsoft.ManagedIdentity/userAssignedIdentities/agentgateway-identity

Note the following details.

• Port 4000 carries LLM traffic. When your configuration file defines no gateway, the implied default gateway serves LLM traffic on port 4000 and MCP traffic on port 3000. Set --target-port to the port that carries the traffic you route. For more information, see Configuration modes.
• The identity is the credential. Because auth.azure uses the identity attached to the container, --user-assigned is what lets agentgateway call Azure OpenAI. No API key is needed in the create command or in the configuration file.
• The container still needs a configuration file. The command above starts the image with no -f flag, so agentgateway generates a default configuration that does not route to Azure. Mount your file with an Azure Files volume, or bake it into your own image, and pass it with -f. For more information, see Configuration storage.
• --ingress internal keeps the gateway inside the environment. A gateway that holds Azure OpenAI access is a credential of its own, so anyone who can reach it can spend against your resource. Before you switch to --ingress external, put an authentication policy in front of it. For more information, see Authentication and identity.

Run on AKS

AKS is an ordinary Kubernetes distribution as far as agentgateway is concerned. Two options are available.

• Run standalone agentgateway as a Deployment with the Helm chart. Enable the workload identity add-on and annotate the pod’s service account, and auth.azure.implicit picks up workload identity automatically.
• Run the Kubernetes control plane, which manages agentgateway proxies from Kubernetes custom resources and the Kubernetes Gateway API.

Role assignments

Assign the roles that agentgateway needs to the managed identity.

# Get the managed identity principal ID
PRINCIPAL_ID=$(az identity show --name agentgateway-identity \
  --resource-group my-rg --query principalId -o tsv)

# Grant Azure OpenAI access
az role assignment create \
  --assignee $PRINCIPAL_ID \
  --role "Cognitive Services OpenAI User" \
  --scope /subscriptions/<sub-id>/resourceGroups/my-rg/providers/Microsoft.CognitiveServices/accounts/my-openai

# Grant Key Vault access, if you store the API keys of other providers there
az role assignment create \
  --assignee $PRINCIPAL_ID \
  --role "Key Vault Secrets User" \
  --scope /subscriptions/<sub-id>/resourceGroups/my-rg/providers/Microsoft.KeyVault/vaults/my-vault

Azure services

Next steps

• Azure for the full provider reference, including Azure AI Foundry.
• Set up the UI to serve the web interface on a gateway.
• Choose where configuration is stored before you mount a read-only file.