← Back to the wire21 Nov 2026
The AI WireDispatch No. 095
foundry-devblog ▸ product· filed 8 Nov 2026 · 2 min read

Microsoft's Foundry extension for Azure Developer CLI reaches 1.0

Microsoft has made its Foundry extension for Azure Developer CLI generally available, reaching version 1.0 with agent setup and local testing. The extension ties agent configuration to a committed azure.yaml file, works with azd provision and azd deploy, and includes an evaluation path that requires explicit review before changes are deployed. It requires Azure Developer CLI 1.37.0 or later, a current uv release, Python 3.13 or newer, and an Azure subscription with Foundry permissions and model quota.

Machine-drafted illustration · reviewed by a humanFIG. 01

Microsoft has made its Foundry extension bundle for Microsoft Azure Developer CLI generally available. The microsoft.foundry extension reaches version 1.0, adding agent setup and local testing to azd, Microsoft's command-line tool for provisioning Azure resources and deploying application code.

Agent configuration lives with the code. A committed azure.yaml file describes the agent, its model deployment, compute allocation, and its dependency on a Foundry project. Because the file is reviewed alongside the source, a pull request can cover both the code and the infrastructure it runs on. The same definition is used for local development and CI/CD pipelines, with each azd environment supplying its own subscription, region, and resource settings.

The workflow is: install the extension, initialize a project from a sample template, run azd provision to create resources, and test locally with Agent Inspector while model calls still hit the Azure deployment. Deployment is handled by azd deploy, which uploads source to Foundry for a remote build, skipping a local container build. If the Azure CLI is already signed in, the sample's DefaultAzureCredential can reuse that login for local model calls.

The extension targets Microsoft's hosted agent concept, where your own agent code runs on managed compute: you choose the framework and orchestration logic, and Foundry handles the runtime. Teams that need custom provisioning can eject editable Bicep or Terraform files, commit them, and continue using azd provision. A separate Microsoft Foundry Skill also lets coding agents such as GitHub Copilot and Claude Code scaffold projects and run local test steps.

An evaluation path is available. A separate extension generates test data and grading rubrics, and azd ai agent optimize can reuse a configured evaluation as a baseline. Optimization candidates still require explicit review before being deployed, so the tooling stops short of auto-promoting changes.

The toolchain has several prerequisites. It requires Azure Developer CLI 1.37.0 or later, a current uv release, and Python 3.13 or newer. An Azure subscription, Foundry permissions, and model quota in the selected region are all required, and provisioned resources and model usage incur Azure charges. Deployment types such as Global Standard or Data Zone Standard vary in availability by subscription and region.

Cleanup behavior depends on whether the Foundry project is new. For a new project, azd down deletes the resource group and everything in it. For a reused project, it leaves the project and its resources, including the hosted agent, in place. A demo environment can therefore outlast the down command if the separate cleanup guidance is not followed.

Read the original at devblogs.microsoft.com →

End of dispatch
More on the wire
99foundry-techcommunity ▸ tooling · filed 19 Nov 2026Deployment Pattern for Foundry Toolboxes with Independent Lifecycles98foundry-devblog ▸ product · filed 10 Nov 2026Microsoft adds agentic mode to Azure Content Understanding for multi-pass document analysis97agent-framework ▸ product · filed 10 Nov 2026Microsoft ships Agent Framework builder skills for coding agents96foundry-techcommunity ▸ product · filed 9 Nov 2026Microsoft Foundry Documents Production Agent Tuning Pipeline Using Agent Optimizer