Weekly .NET Roundup - Agentic Blazor UI and IDE AI Workflows

This week's .NET roundup centers on practical “agentic” app patterns, with experimental Blazor AI components in .NET 11 RC1 that treat tool calls, approvals, and streamed output as first-class UI building blocks. The IDE story keeps moving from chat help to task-driven workflows, as Visual Studio adds BYOM for Copilot agent mode, NuGet Audit fixes from the Error List, and MCP-enabled Git agent features, while VS Code improves multi-folder and session controls that matter for multi-repo .NET work. On the Azure side, connector-based triggers for .NET isolated-worker Functions and SDK updates (including stable Entra ID auth for PostgreSQL and AI Search preview APIs) aim to reduce glue code and make integrations feel native.

This Week's Overview

Blazor's new “Agentic UI” direction with AI components

Building on last week's shift from isolated AI demos toward full agent workflows (MAF orchestration, MCP tool calling, and Blazor UIs like Interview Coach), Blazor picked up a concrete set of building blocks for “agentic UI” in the experimental Blazor AI components shipping with .NET 11 RC1. Instead of treating AI as a single chat panel, the new components are designed for workflows where an agent can stream content into the page, surface tool calls as UI, and request explicit approvals before acting.

The .NET team is leaning into patterns you can compose across the app: streamed content blocks for incremental responses, tool rendering so actions show up as first-class UI elements, and shared or predictive state patterns so the UI can stay responsive while the agent reasons. The end-to-end sample ties together Microsoft.Extensions.AI, AG-UI, Microsoft Agent Framework (MAF), ASP.NET Core, Foundry, and .NET Aspire, which makes it easier to see how the UI layer fits into a broader agent stack.

If you already build Blazor line-of-business apps, the practical takeaway is that the UI can now model “decision points” (approvals) and “work steps” (tool calls) without custom plumbing for each agent integration. Because the components are experimental in RC1, plan for API churn and keep the agent-facing UI isolated behind your own wrappers until the patterns settle.

Visual Studio and VS Code updates center on Copilot, agent workflows, and day-to-day debugging

Following last week's editor focus on making agent sessions more governable and debuggable inside VS Code, both Visual Studio and VS Code shipped updates that keep pushing AI assistance from “chat help” into deeper, task-driven workflows. This week connected the dots between new Copilot agent capabilities, better controls for multi-repo work, and a few quality-of-life improvements that matter when you're in the debugger or working with containers.

Visual Studio September Update: BYOM for agent mode, NuGet Audit fixes, MCP-enabled Git agent, and more

Building on last week's recurring theme of MCP as the practical “tool contract” for production agent workflows, Visual Studio's September update added “Bring Your Own Model” (BYOM) for Copilot agent mode, letting teams point the agent experience at a model that fits their needs (for example, enterprise constraints or specialized quality requirements). It also brings Copilot-assisted fixes for NuGet Audit warnings directly from the Error List, turning dependency security warnings into something you can act on without leaving the IDE.

On the workflow side, a new Git agent focuses on pull request exploration, and it includes support for MCP server integration (MCP is increasingly used as a standard way for agents to call tools and access resources). There are also targeted developer productivity improvements like better C# debugging for compound if conditions and the ability to attach to processes running in Podman, which helps teams standardizing on container workflows outside of Docker Desktop.

VS Code 1.140 and release recap: Copilot iteration plus multi-folder and session controls

Picking up from last week's “agent-first development” walkthrough in VS Code (where session controls and the chat debug view were central), VS Code 1.140 highlighted continued Copilot iteration, including a feature called HydraFusion (as demoed in the walkthrough) and improvements aimed at how people actually work across repos. Multi-folder sessions and tighter session composer controls are a practical upgrade for .NET developers juggling microservices, shared libraries, and infra repos in one editor window.

The broader “Release Recap” stream reinforced that VS Code's recent cadence is about pairing editor capabilities with Copilot Chat and related tooling, rather than treating Copilot as a separate mode. If your team uses VS Code as the default .NET editor, the key action item is to review session and multi-folder behaviors in your workspace setup, since those changes can affect how tasks, terminals, and Copilot context behave across folders.

Azure's .NET story this week was about reducing the glue code needed to integrate with platform services. Between new connector-based triggers and incremental SDK releases, the common theme is making integrations feel more “native” inside .NET apps and Functions.

Azure Connectors SDK (preview): bring Logic Apps connectors into .NET 10 isolated-worker Functions

The Azure Connectors SDK entered preview with a model that lets you use Azure Logic Apps-style connectors directly in .NET applications. The walkthrough example builds a .NET 10 isolated-worker Azure Functions app that triggers on new Office 365 email (via the Office 365 Outlook connector) and sends a formatted notification, using a Connector Namespace and explicit trigger registration.

Security and operational details are part of the story here: the sample covers managed identity access policies so the function can authenticate to the connector without embedding credentials. For teams already invested in Functions, this approach can replace custom polling code or bespoke integration layers, but you should treat it as a preview surface and validate connector limits, latency, and deployment requirements before relying on it for production flows.

Azure SDK Release (September 2026): AI Search preview APIs and Entra ID auth for PostgreSQL in .NET

The September 2026 Azure SDK release roundup called out Azure AI Search preview API support across languages, which matters if you're building RAG (retrieval-augmented generation) or hybrid search workloads and need access to the newest service capabilities. On the .NET side, a particularly practical update is stable Microsoft Entra ID authentication libraries for PostgreSQL, which can simplify passwordless database access patterns.

If you're standardizing on managed identity or Entra-backed auth in production, having stable client libraries reduces the need for custom token plumbing and makes it easier to align with enterprise security policies. The post also notes ecosystem improvements like a stable Rust Azure Storage SAS crate, which is relevant if your organization uses multiple languages around a .NET core.

Other .NET News

Azure Policy Linter became open source, giving teams a straightforward way to enforce Azure Policy authoring quality earlier in the lifecycle. You can install it as a .NET global tool, run it against one or many policy definition files, and use JSON output and rule sets to plug policy linting into CI/CD so common mistakes get caught before deployment.