Introducing azd AI Agent Extension for Seamless Foundry Agent Publishing

Microsoft unveils a game-changing extension for Azure Developer CLI (azd) at Ignite 2025, promising to streamline the creation, provisioning, and deployment of AI agents making AI development faster, easier, and more secure.

What Is the New azd AI Agent Extension?

Announced at Ignite 2025, the azd AI agent extension aims to transform how developers build AI agents on Microsoft Foundry. It enables a unified, streamlined experience that bridges local development with cloud deployment—allowing developers to create, provision, and publish agents directly from their terminal or code editor.

This extension integrates foundational features of Foundry—such as multi-model reasoning, evaluation tools, and model deployment—into the familiar azd workflow, empowering rapid iteration and reliable cloud deployment with minimum friction.

Key Features of the azd AI Agent Extension

1. Effortless Project Initialization

Developers can now start an AI agent project using predefined templates that include all necessary Infrastructure as Code (IaC) files. This allows for a quick setup—creating directory structures, configuration files, and IaC scripts automatically.

2. Declarative Infrastructure Management

All resources, services, and models are defined in a single azure.yaml configuration file. This declarative approach makes infrastructure easy to version control, share across teams, and replicate in different environments.

3. Unified Provisioning and Deployment

With a simple azd up command, developers can provision resources, deploy models, build container images, and publish their agents—all in one seamless operation. This automation accelerates deployment times from hours to minutes.

4. Dynamic Agent Definition Management

Pull existing agent definitions from GitHub or local repositories, analyze them automatically, and update project configurations with minimal manual effort.

5. Built-in Security

The extension automatically configures security best practices—using managed identities, role assignments, and Azure AD authentication—without needing manual setup. This ensures your agents are secure by default.

How to Get Started with azd AI Agents

Getting your first AI agent up and running on Foundry is straightforward. Microsoft recommends a simple three-step process:

Prerequisites

Install the latest Azure Developer CLI ( azd ) (version 1.21.3+).
Install the azd AI agent extension ( azd extension install azure.ai.agents ).
Have an active Azure subscription with permissions to create resources.
Install Azure CLI ( az ) for certain operations.

Step 1: Initialize a Project Using a Template

Create a new project swiftly with the basic setup:

  
    azd init -t Azure-Samples/azd-ai-starter-basic

Provide a project name, and the CLI will automatically set up the necessary folder structure and configuration files.

Step 2: Define Your AI Agent

Download an agent definition (e.g., a calculator agent) from GitHub or provide your own:

  
    azd ai agent init -m <agent-definition-url>

This command fetches, analyzes, and maps your agent's requirements into the project configuration.

Step 3: Deploy With a Single Command

Deploy the entire agent setup to Azure:

  
    azd up

Within minutes, your agent is live, hosted on Foundry, and accessible via a secure endpoint—ready for integration and testing.

Behind the Scenes: How It Works

The extension automates complex processes such as infrastructure provisioning, containerization, model deployment, and security configuration:

Project scaffolding sets up Bicep files, environment configs, and agent definitions.
Resource provisioning automates resource creation, model deployment, and environment setup based on declarative configs.
Container management packages custom code into containers, pushes to Azure Container Registry, and deploys agents seamlessly.
Security configurations are handled automatically, ensuring robust, best-practice security practices with managed identities and Azure AD authentication.

Use Cases and Scenarios

This powerful extension opens up numerous possibilities for developers and organizations:

Conversational AI assistants : Deploy chatbots capable of contextual understanding, knowledge retrieval, and personalized responses.
Data analysis agents : Automate insights, visualizations, and complex calculations with specialized models.
Multi-agent systems : Coordinate multiple agents for complex workflows, orchestration, and enterprise-wide automation.
Enterprise deployment : Standardize agent development, enforce security, and integrate with CI/CD pipelines for seamless scaling and management.

Customization for Advanced Needs

Developers with more experience can fine-tune deployments by editing the azure.yaml file to specify models, resources, capacities, and environment variables, allowing for tailored, high-performance agents.

Get Started Today

The azd AI agent extension is now in public preview , giving developers access to cutting-edge tools that simplify AI deployment workflows. Whether you're building customer service agents, data analysis tools, or multi-agent systems, this extension aims to make AI development faster, more manageable, and more secure.