AI Toolkit for VS Code Gets Major Upgrade with Unified Agent Development Experience

Microsoft has released version 0.32.0 of the AI Toolkit for VS Code, bringing significant enhancements designed to streamline the development of production-ready AI agents. This update represents a major step forward in unifying the development experience across AI Toolkit and the Microsoft Foundry extension.

Unified User Experience Across Extensions

One of the most notable changes in this release addresses user feedback about navigating between AI Toolkit and Microsoft Foundry extensions. Microsoft has consolidated the user experience by merging the Foundry sidebar directly into AI Toolkit, creating a more cohesive workflow.

The new My Resources view unifies the AI Toolkit and Foundry extension sidebar panels into a single interface. Local resources such as models, agents, and tools are now grouped under a Local Resources node, while Foundry remote resources appear alongside them. All nodes are collapsed by default and maintain their state across sessions, reducing visual clutter and improving navigation.

Additionally, the Developer Tools panel now supports two switchable layouts accessible from the panel title menu: Group by Lifecycle (organizing entries into Discover, Build, and Monitor stages) or Group by Resource (grouping them into Agent Dev Tools and Model Tools). Foundry agents and models now feature dedicated icons to distinguish them from local resources, and Foundry Model Licenses are displayed directly in the model catalog.

Enhanced Agent Creation Experience

The update introduces a Create Agent View that serves as a unified entry point for creating AI agents. This new interface offers two distinct paths side by side: users can either create agents in code with full control by scaffolding projects from templates or generating agents using GitHub Copilot, or they can design agents without code by launching Agent Builder directly through the UI.

GitHub Copilot Integration and Skill-Based Development

A significant enhancement involves GitHub Copilot agent development. Agent code generation, evaluation, and deployment now utilize the open-source Microsoft foundry skill from GitHub—the same source used by GitHub Copilot for Azure. AI Toolkit automatically installs and maintains this skill, eliminating manual setup requirements.

Microsoft emphasizes the importance of building with skills, optimizing the toolkit to allow seamless use of skills within workflows. This deep Copilot integration enables developers to build agents directly within their development environment, taking full advantage of the enhanced capabilities.

Agent Builder Improvements

Agent Builder continues to evolve with major usability and functional upgrades. A dedicated Conversations tab is now available for Foundry agents, making it easier to review and manage conversation history during development. The addition of auto-save functionality ensures that draft agents are automatically saved before running in the playground, preventing accidental loss of unsaved changes.

Integration with Foundry has been enhanced, allowing users to open Foundry prompt agents directly in Agent Builder from the Foundry extension. The Foundry Prompt Optimizer enables generation and refinement of agent instructions. Users can now configure MCP tool approval settings, choosing between auto or manual approval for MCP tool calls during agent runs, providing complete control over tool invocation handling.

The View Code support for workspace scaffolding allows developers to quickly generate the project structure needed for Foundry agents, streamlining the initial setup process.

Updated Evaluation Framework

The evaluation framework has been updated to use the pytest-agent-evals SDK for running agent evaluations. This update aligns perfectly with the latest evaluation framework, providing developers with more robust testing capabilities for their AI agents.

Getting Started

Developers can download the AI Toolkit from the Visual Studio Code Marketplace and explore comprehensive AI Toolkit Documentation to get started with the new features.

Microsoft encourages users to provide feedback through their GitHub repository, welcoming feature requests, bug reports, and general feedback on the development experience.

This update positions AI Toolkit as a more powerful and unified platform for AI agent development, addressing key pain points in the workflow and providing developers with enhanced tools for building production-ready AI agents directly within VS Code.