With Azure AI Foundry • Enkaidu

Start with Azure AI Foundry

New in 0.9.13 Microsoft AI Foundry (also referred to as Azure AI Foundry) exposes Open AI / chat-completion models, which Enkaidu can drive through the new azure_foundry provider.

If you have an Azure AI Foundry project and API key, create a file enkaidu.yaml and copy in the following configuration, then run enkaidu from the command line in the same folder as the config file.

session:
  model: az_gpt5_6luna
auto_load:
  toolsets:
     - DateAndTime
llms:
  azure-ai:
    provider: azure_foundry
    env:
      AZURE_FOUNDRY_ENDPOINT: $AZURE_FOUNDRY_ENDPOINT
    models:
       - name: az_gpt5_6luna
        model: gpt-5.6-luna

If you’re using a different model, change the values pointed to by the name: / model: pair (the name is what you reference in session: model:, the model is the name as defined by your Foundry project).

Environment variables

To use an Azure AI Foundry model, you will need to set the following environment variables before running Enkaidu.

Env. variable	Need	Description
`AZURE_FOUNDRY_API_KEY`	Required	The API key for your Foundry project
`AZURE_FOUNDRY_ENDPOINT`	Required	The bare endpoint URI for your project, e.g. `https://myproject.services.ai.azure.com`. Just the endpoint, no paths

AZURE_FOUNDRY_ENDPOINT should be the project’s bare base URI only (no version or chat path) — for example https://myproject.services.ai.azure.com, not a URL with a trailing path.