Microsoft has rolled out a major update to the AI Toolkit for VS Code, and it’s clearly pushing agent development closer to a normal software dev workflow.
The update introduces a Tool Catalog for discovering and managing agent tools, a new Agent Inspector that brings F5 debugging and step-through execution to agents, and “Evaluation as Tests,” which lets developers run AI evaluations like standard test cases.
The big theme: treating agents less like black boxes and more like software you can debug, test, and manage directly inside the IDE.
Full article here: Microsoft Sharpens AI Toolkit for VS Code in Foundry Update - Visual Studio Magazine
What do you think matters more for agent development right now: better tools or better models?