Azure Weekly

Issue 567

14th June 2026

A quieter week after the deluge of BUILD announcements; the edition contains plenty of community content! The big news this week was that Fable 5 was released by Anthropic, only to be suspended by the US Government a few days later. I did manage to use it for a couple of days to tackle some complex coding problems that Opus struggled to complete. I was very impressed by the results, and I hope the issues are resolved soon.

Highlights this week include:

  • ACI Sandboxes: A Glimpse into the Future of Agentic Compute by Dana Cozmei - ACI Sandboxes introduce Direct Virtualization (L1VH) to combine Hyper-V security with container agility, enabling isolated AI agent environments that can launch 10,000 instances in two seconds for secure, scalable agentic compute.
  • Generally Available: SQL MCP Server by The Azure Updates Team - SQL MCP Server offers secure, high‑performance access to production databases for building agentic applications, supporting both PostgreSQL and Azure Cosmos DB seamlessly.
  • What's new with Postgres at Microsoft, 2026 edition by Claire Giordano - The 2026 update covers major advancements in Azure Database for PostgreSQL flexible server, including Premium SSDv2 storage enhancements for better performance and cost efficiency, as well as new infra‑as‑code deployment options for Elastic Clusters using Terraform, Bicep, and Ansible.
  • Stop wasting time and use Custom Extensions for PIM approvals by Daniel Bradley - Custom extensions in Entra PIM let you automate approval workflows, automatically validate role activations and integrate with ticketing systems to cut manual administrative tasks by up to 40%.
  • What's new in Azure App Service at #MSBuild 2026 by Mayunk Jain - Azure App Service’s #MSBuild 2026 launch introduces an Easy AI experience that lets web apps become AI‑ready instantly, along with Isolated v4 for enhanced security and compliance, Managed Instance for seamless legacy migration, and platform CLI upgrades to streamline deployments—accelerating AI application development.
  • When AI Automates the Joy Out of Work by Adam Bertram - The article explores how AI-driven automation enhances productivity but may diminish the intrinsic satisfaction found in manual technical tasks.

Finally, we're publishing a series of blog posts about the latest version of our high performance, zero allocation JSONSchema library:

  • Introducing Corvus.Text.Json V5: Why V5 Exists by Matthew Adams - Corvus.Text.Json V5 introduces a mutable builder model and pooled memory architecture to dramatically reduce allocation overhead and improve performance in high-throughput .NET applications that process large volumes of JSON data, offering distinct trade‑offs from the immutable V4 engine.

🤖 AI

  • Claude Fable 5 available today in Microsoft Foundry: Powering the next era of autonomous agents Claude Fable 5, Anthropic’s newest frontier model, is now accessible via Microsoft Foundry, enabling autonomous agents to tackle complex tasks like code refactoring and deep research synthesis while incorporating strong safety measures for broad enterprise use.
  • Available today: Anthropic Claude Fable 5 in Microsoft 365 Copilot Claude Fable 5 in Microsoft 365 Copilot supports longer, multi-step workflows and integrates Work IQ for context‑aware planning, drafting, and iteration across complex tasks within an enterprise‑ready framework.
  • Azure OpenAI Architecture: The Decisions That Actually Matter (Part 1) Azure OpenAI’s production architecture hinges on five critical design decisions—capacity model, deployment location, governance layer, grounding strategy, and quota engineering—that determine resilience under variable loads, compliance constraints, and evolving model versions.
  • Azure OpenAI Architecture: The Decisions That Actually Matter (Part 2) Azure OpenAI Architecture Part 2 explains how to apply the Azure Well‑Architected Framework's five pillars—Cost Optimization, Security, Reliability, Performance Efficiency, and Operational Excellence—to GenAI workloads, offering best practices for managing model upgrades, Priority Processing trade‑offs, and building a
  • Creating a multi-agent application - Part 1 Creating a multi-agent application in C# involves designing autonomous agents that collaborate to enhance functionality while requiring careful orchestration to manage complexity and debugging challenges.
  • Creating a multi-agent application - Part 2 This Python Colab notebook demonstrates building a multi-agent application using LangChain and LangGraph, featuring a Blogger orchestrator, Researcher for web searches, Author to draft posts, and Reviewer for feedback, all integrated with OpenAI's gpt-4o-mini model and Tavily search tool
  • Creating a multi-agent application - Part 3 This article continues a multi-agent application series by detailing how to define and implement a Research agent template, using Tavily search for data retrieval and an LLM for summarization, following a consistent pattern seen with previous agents like Blogger.
  • The Runtime Is Free Now. The Harness Is the Business. Microsoft has turned its formerly restricted open‑source agent runtime OpenClaw into a commodity, focusing instead on standardizing and monetizing the harness that provides identity, access control, approval gates, and data protection features essential for enterprise compliance.
  • Building my own AI cycling coach with Claude and Strava Elio Struyf created a custom Claude skill that integrates with the Strava Connector to act as an AI cycling coach, personalizing weekly plans around social rides and strength training based on user preferences and recent activity data.
  • Now in Foundry: Command A+ (W4A4), Chandra OCR 2, and GLM-OCR Hugging Face’s Foundry now hosts Cohere Labs’ Command A+ (W4A4), a 218B‑parameter low‑bit quantized reasoning model optimized for agentic tasks across 48 languages; Datalab’s Chandra OCR 2, a 5.3B
  • Transitioning from Azure Language Features to Foundry Models Transitioning from Azure Language Features to Foundry Models involves retiring eight capabilities—including Key Phrase Extraction, Entity Linking, and Summarization—by March 2029 (with Entity Linking ending earlier), encouraging teams to adopt Foundry‑based models like Mistral‑Small‑2503 for
  • Turn specs into evals for any agent with ASSERT ASSERT turns natural‑language behavior specs into full evaluation pipelines, automatically generating test scenarios, datasets, metrics, and scorecards to ensure AI systems meet product‑specific requirements efficiently and stay up‑to‑date with evolving policies and tools.
  • A new chapter of efficient foundation models for medical imaging Microsoft introduces premium, fully managed medical imaging AI models on Foundry, addressing customer needs for scalable, HIPAA‑compliant endpoints and simplified fine‑tuning to support clinical workflows without managing infrastructure.
  • How Machine Learning Fails with Megan Robertson Megan Robertson emphasizes setting clear expectations and metrics in ML projects to avoid pitfalls like overfitting test data, stressing the critical role of high-quality datasets and validation techniques for reliable results.
  • Driving Stop-and-Go Business Processes to Closure with Foundry Hosted Agents Foundry Hosted Agents enable AI agents to maintain long-term state across weeks of work using per-project sandbox VMs, eliminating the need for external databases and simplifying session handling for complex business workflows like Statement of Work response orchestration.
  • When AI Automates the Joy Out of Work The article explores how AI-driven automation enhances productivity but may diminish the intrinsic satisfaction found in manual technical tasks.

🔎 Analytics

🖥️ Compute

🚢 Containers

  • ACI Sandboxes: A Glimpse into the Future of Agentic Compute ACI Sandboxes introduce Direct Virtualization (L1VH) to combine Hyper-V security with container agility, enabling isolated AI agent environments that can launch 10,000 instances in two seconds for secure, scalable agentic compute.
  • AKS MaxUnavailable Fallback Now in Preview The AKS MaxUnavailable fallback feature in preview allows clusters to automatically adjust pod availability during node failures, reducing downtime and improving resilience without manual intervention.
  • Designing for High Availability: The Operational Reference for Running a Geo-Replicated ACR Designing for High Availability explains how Azure Container Registry’s health‑aware failover automatically reroutes traffic during regional incidents without manual intervention, detailing authentication, throttling management, and endpoint configuration best practices.
  • Troubleshooting ML Model Loading, GPU Issues, and Memory Pressure in Azure Container Apps Azure Container Apps introduces specialized troubleshooting guidance for ML model loading, GPU resource contention, and memory pressure issues that arise during AI workload deployment, offering step‑by‑step fixes such as separating startup probes from liveness checks to prevent premature container termination.
  • Performance Tuning Cold Starts, Scaling Delays, and Startup Latency in Azure Container Apps Azure Container Apps’ cold start latency can be reduced from tens of seconds by setting minReplicas to 1 and optimizing image sizes or configuring custom readiness probes.
  • AI 200 – Azure Container Apps Express: Blazing-Fast Deployments Without the Overhead Azure Container Apps Express (Preview) streamlines container deployment by eliminating pre-provisioned environments and reducing required fields, enabling developers to launch a publicly accessible app in seconds via the dedicated portal or Azure CLI.
  • Troubleshooting Azure Container Apps and Jobs for .NET and Django Workloads Azure Container Apps and Jobs for .NET and Django workloads often encounter restart loops due to startup crashes, misconfigured environment variables or secrets, and health‑probe failures that cause endless container restarts. This article explains the root causes—such as unhandled exceptions from missing database contexts, incorrect connection strings
  • 3 things leaders need to know from Microsoft Build 2026 At Build 2026, Microsoft emphasized that successful AI adoption hinges on integrating models with a shared intelligence foundation—like the new IQ layers—that give every AI system context about your business, ensuring tools evolve into scalable systems capable of delivering real outcomes.
  • KubeCon + CloudNativeCon Europe 2026: Reflections from the Field KubeCon + CloudNativeCon Europe 2026 offered actionable insights and fostered collaboration, making it a pivotal event for deepening Kubernetes expertise and community connections.
  • We Gave Ourselves 20 Minutes to Build an AI Agent for a Lumber Company. The Timer's Still on Screen. The webinar demonstrates how to build and deploy an AI agent for a lumber company's inventory decision problem using Azure Foundry in just 20 minutes, showcasing real-time recommendations, guardrails, and a fully functional chat application with five simple commands.
  • Dapr v1.18 is now available Dapr 1.18 introduces enhanced security features, including optional cryptographic signing of workflow history for tamper detection and a new WorkflowAccessPolicy resource to enforce granular access control across applications.
  • Building Dapr Workflows in .NET With Aspire Dapr Workflows in .NET with Aspire enable deterministic process orchestration, ensuring durability and seamless replayability across crashes or deployments while integrating smoothly with Redis Iris for real-time context management.
  • Running LiteLLM on AKS with azd and Bicep Running LiteLLM on AKS with azd and Bicep enables a fully managed self‑hosted LLM gateway that includes private networking, Redis caching, PostgreSQL for spend tracking, and automatic TLS ingress, simplifying authentication, rate limiting, and budget management across multiple LLM providers.
  • Deploying Foundry Hosted Agents from Source Code Foundry now allows Hosted Agents to be deployed directly from .zip files, reducing complexity for customers who prefer source‑code over containerized workloads.

🗄️ Databases

  • Generally Available: SQL MCP Server SQL MCP Server offers secure, high‑performance access to production databases for building agentic applications, supporting both PostgreSQL and Azure Cosmos DB seamlessly.
  • Generally Available: Microsoft Entra server principals on Azure SQL Database The GA release of server principals in Azure SQL Database lets you map Microsoft Entra identities to SQL logins using CREATE LOGIN ... FROM EXTERNAL PROVIDER, aligning authentication capabilities with on-premises SQL Server.
  • Generally Available: PostgreSQL Hub for Azure Developers PostgreSQL Hub for Azure Developers offers curated sample applications, solution accelerators, and step‑by‑step learning paths to accelerate AI and app development on Azure PostgreSQL instances.
  • Generally Available: Maintenance control for Azure Database for PostgreSQL - reschedule, apply on demand, view and download Azure Database for PostgreSQL Flexible Server now offers general availability of maintenance control features, enabling users to reschedule, apply updates on demand, and view or download maintenance logs to minimize service disruptions.
  • Retirement: Azure Synapse Link for Azure Cosmos DB NoSQL Azure will retire the Synapse Link feature for Cosmos DB NoSQL on March 31, 2026, with full support for current users extending to March 31, 2029, before the service officially ends of life.
  • What's new with Postgres at Microsoft, 2026 edition The 2026 update covers major advancements in Azure Database for PostgreSQL flexible server, including Premium SSDv2 storage enhancements for better performance and cost efficiency, as well as new infra‑as‑code deployment options for Elastic Clusters using Terraform, Bicep, and Ansible.
  • Introducing Durable Functions in PostgreSQL pg_durable, an open‑source PostgreSQL extension developed by Azure engineers, enables long‑lived, fault‑tolerant workflows directly within Postgres, offering built‑in retries, parallelism, scheduling, and recovery without external orchestration.
  • Introducing Azure HorizonDB - PostgreSQL Azure HorizonDB extends PostgreSQL with built‑in AI capabilities, including real‑time vector search via DiskANN and AI re-ranking, while maintaining compatibility with existing SQL workloads.
  • Our MongoDB TLA+ Workshop The second MongoDB TLA+ workshop incorporated AI tools like Claude Code, dramatically increasing conversion rates and enabling nearly all attendees to start modeling real work projects within minutes.
  • Common Cache Strategies in Azure Managed Redis Azure Managed Redis offers four cache strategies—Data Cache to store frequent DB query results, Content Cache for static page fragments, Session Store for user sessions—helping applications cut latency by up to 95% compared with direct database access.
  • SSIS 2025 Microsoft.Data.SqlClient Connection Error The article explains how Anthropic's Fable 5 model helped diagnose and resolve an SSIS connection error by identifying a mismatch between referenced and installed versions of Microsoft.Data.SqlClient, then correcting a TLS 1.0 compatibility issue with Azure SQL DB using registry settings.
  • Azure Cache for Redis vs. Enterprise vs. Azure Managed Redis: What Retires When, and Where to Go Migrating from Azure Cache for Redis to its replacement, Azure Managed Redis (AMR), involves updating connection strings from port 6380 to 10000 and adopting new tiers designed for higher performance and lower operational overhead.
  • SQL ART: Who's Blocking Who? Visualising SQL Server Blocking With Spatial Geometry SQL ART introduces a live blocking visualizer using SQL Server’s geometry data type to chart wait times, blocked sessions, and head blockers directly in SSMS, making complex blocking issues instantly understandable for DBAs.
  • How data scientists work with PostgreSQL from Python (and how to write queries that don’t break at scale) The article explains how data scientists leverage Python libraries like psycopg2 and SQLAlchemy to interact with PostgreSQL, offering best practices for writing scalable queries that avoid performance pitfalls at large dataset sizes.
  • AI-200 Developers – What Is Azure Managed Redis Azure Managed Redis offers enterprise‑grade Redis Enterprise capabilities—including native vector search via RediSearch, active‑active geo‑replication, up to 13 TB storage on supported tiers, and Entra ID authentication—making it the recommended in‑memory data platform for AI‑200 workloads
  • Build a Knowledge Graph in Azure HorizonDB with AI Functions and Apache AGE Building a knowledge graph in Azure HorizonDB with AI Functions and Apache AGE enables answering complex cross‑service incident questions like upstream failure causes, team resolution rates, and cascading failure timelines with single Cypher queries.
  • Database AI Agents: The Read-Only Rule The read‑only advisory system in Database AI Agents monitors databases and flags suspicious states without making any changes, ensuring data integrity while providing actionable insights.
  • It’s 2026. Why are databases still failing GDPR compliance audits? The article reveals that persistent GDPR compliance failures stem largely from technical database issues like complex erasure across relational schemas, backup paradoxes undoing deletions, and audit log contradictions, highlighting gaps in SQL Server, PostgreSQL, Oracle, and MySQL implementations.
  • And Then There Was The Time RCSI Actually Made Query Results More Accurate. The article demonstrates how switching to Read Committed Snapshot Isolation (RCSI) resolved inaccurate query results caused by row locking under default pessimistic isolation levels in SQL Server.
  • Dynamic SQL vs OR Clauses in SQL Server This video explores using dynamic SQL to optimize queries with OR clauses in SQL Server, offering performance benefits over traditional case expressions and join clauses for complex search scenarios.
  • Announcing the Release of SSMS 22.7.0 - and many previews! SSMS 22.7.0 introduces a "What’s New" page for quick release highlights, T-SQL formatting preview with customizable styling options, and Agent mode for GitHub Copilot to handle complex multi-step tasks in SQL development workflows.
  • Remigration – T-SQL Tuesday #199 Remigration—moving back from the cloud to on‑premises or private data centers—affects SQL Server Integration Services (SSIS) by increasing package executions up to 15% due to economic factors, while also boosting demand for related training resources.
  • Make upgrades predictable with Pre-Upgrade Validation Checks Pre-Upgrade Validation Checks for Azure Database for PostgreSQL let you identify and resolve compatibility issues before initiating major version upgrades, reducing failed attempts and improving production readiness.
  • Moving Indexes To A New Filegroup: Microsoft Still Hates You Moving indexes to a new filegroup in SQL Server is notoriously cumbersome—especially when dealing with heaps, nonclustered indexes containing large objects (LOBs), or simply rebuilding normal indexes requires scripting and recreating them on the target filegroup without an easy “move” option.
  • How to call an Ollama-based AI text embeddings model from SQL Server 2025 This article shows how to integrate Ollama‑hosted AI text embeddings models with SQL Server 2025 by enabling REST endpoints, defining an external model via CREATE EXTERNAL MODEL, and calling AI_GENERATE_EMBEDDINGS in T‑SQL for clean, reusable vector generation.
  • A Case for Simulation-Driven Resilience in Agentic Data Systems Simulation‑driven resilience addresses metastability by using MESSI to model distributed system interactions, uncovering hidden failure modes before they cause production incidents.
  • Vector Search in SQL Server 2025: Storing Embeddings, Querying Them, and What to Watch Out For SQL Server 2025 introduces GA vector storage for embeddings, allowing semantic search on existing relational data without separate databases, with performance considerations for production-scale embedding dimensions.

🛠️ Developer tools

🔩 DevOps

  • Azure Monitor SLIs now Generally Available Azure Monitor’s new Service Level Indicators (SLIs) enable organizations to gauge application performance from the end‑user perspective, complementing traditional infrastructure monitoring with metrics like availability and latency that directly reflect customer experience.
  • Azure Monitor Metrics Export Generally Available Azure Monitor Metrics Export’s general availability introduces data collection rules for scalable, low-latency export of multidimensional metrics to storage, Event Hubs, or Log Analytics, now supporting 44 regions with improved control over exported data.
  • Announcing the Path to Production for Agents Webinar Series The Path to Production for Agents webinar series offers a two‑day deep dive on operationalizing AI agents, covering governance, architecture, security, observability, and cost optimization with expert-led sessions and practical guidance.
  • GitHub availability report: May 2026 GitHub's monthly availability report for May 2026 details nine incidents causing service disruptions, with a major pull requests outage lasting over an hour due to a database migration overload, and outlines ongoing infrastructure improvements like moving traffic to Azure, isolating services, and enhancing monitoring to prevent future issues.
  • How AI reduce IT team size & where the impact comes from AI's biggest impact isn't just automating tasks but streamlining handovers, cutting coordination delays, allowing specialized roles to merge into broader outcome-focused positions.
  • June Patches for Azure DevOps Server These June patches for Azure DevOps Server include critical security updates and performance improvements, with the most recent version receiving widespread attention for its enhanced reliability in self‑hosted deployments.
  • Copilot Autofix for GitHub Advanced Security for Azure DevOps Copilot Autofix for GitHub Advanced Security in Azure DevOps uses AI to automatically generate fixes for detected vulnerabilities, streamlining remediation within the same development workflow.
  • Review pull requests without leaving Visual Studio Visual Studio now lets developers review, comment on, and approve GitHub and Azure DevOps pull requests entirely within the IDE, streamlining collaboration without switching browsers.
  • IT Teams evolution in AI era The article explains how AI is transforming IT team structures by reducing handovers between roles, leading to broader responsibilities like Product Discovery Lead, Delivery Manager, and AI-native Solution Architects, ultimately streamlining workflows and improving efficiency.
  • Azure DevOps Pipeline cache issue stopping Dependency Check task The Azure DevOps Pipeline cache issue causing the Dependency Check task failure is due to outdated installer files, which can be resolved by manually updating the installer path and ensuring the pipeline runs with sufficient permissions.
  • Copilot Code Reviews for Azure Repos Microsoft is launching a limited public preview of GitHub Copilot code reviews directly within Azure Repos pull requests, allowing organizations still using Azure Repos to leverage AI‑powered review capabilities without migrating to GitHub.
  • Writing Code vs. Shipping Code: Productivity Effects Across Generations of AI Coding Tools The MIT and Wharton study finds that while AI coding tools boost commit activity by up to 180%, their productivity gains diminish significantly as work progresses toward production, with only modest increases in shipped software releases across three generational tiers of tool adoption.
  • Connect Metrics to Traces with Exemplars in Azure Monitor Azure Monitor now integrates Exemplars to link OpenTelemetry metrics with traces, enabling engineers to quickly trace specific metric spikes back to their root causes and improve troubleshooting efficiency in cloud-native applications.
  • Enterprise Live Migrations: Moving from Azure DevOps Repo to GitHub with minimal disruption Enterprise Live Migrations enables Azure DevOps teams to transition repositories to GitHub with near-zero downtime, using continuous sync and a brief final cut‑over window.
  • GitHub for Beginners: Answers to some common questions This guide explains how to set up SSH keys for secure GitHub authentication and create a Personal Access Token (PAT) for programmatic access, detailing step‑by‑step commands and best practices for managing both methods.
  • Azure Boards vs Jira Azure Boards offers seamless integration within the Azure DevOps suite, providing tight coupling between work items and source control, CI/CD pipelines, and testing tools, ideal for organizations prioritizing a unified development ecosystem, while Jira excels in modular flexibility, allowing extensive customization across diverse third-party tools through its

🧬 Hybrid + multicloud

🎭 Identity

  • Public Preview: Azure Managed Redis now supports Entra ID based RBAC for data management Azure Managed Redis public preview introduces Entra ID based RBAC, enabling fine-grained control over read, write, and administrative access to Redis data without shared keys.
  • Docker sandbox credential injection for Business Central Docker’s new sandbox credential injection feature securely authenticates Business Central environments using Windows authentication with NTLM credentials managed by a host proxy, eliminating the need for manual login within the sandbox while addressing certificate revocation list validation issues.
  • Agent 365 | Identity & Access Controls in Entra Agent 365 enables organizations to manage AI agents across multiple platforms using a unified registry, assign unique Agent IDs via CLI or SDK, enforce least-privilege access through Conditional Access policies and Agent Blueprints, and gain full audit visibility into agent activity with sign‑in logs.
  • Stop wasting time and use Custom Extensions for PIM approvals Custom extensions in Entra PIM let you automate approval workflows, automatically validate role activations and integrate with ticketing systems to cut manual administrative tasks by up to 40%.
  • Fix invalid Microsoft service API permissions in SharePoint The article explains how to resolve an error approving Azure Storage API permissions in a SharePoint Framework web part by manually creating the missing Service Principal for the Azure Storage service using the Azure CLI, due to the storage service not being automatically provisioned in tenants without an active Azure subscription.
  • How to Stop Guests From Inviting More Guests in Microsoft Entra This guide shows how to configure Microsoft Entra external collaboration settings to prevent guests from inviting additional users, enhancing tenant security.
  • Connect-Entra is not recognized The article provides a detailed, step‑by‑step PowerShell guide to resolve the “Connect‑Entra is not recognized” error by installing and importing the official `Microsoft.Entra` module, ensuring seamless identity automation.
  • Connecting to Azure DocumentDB (with MongoDB compatibility) using Microsoft Entra ID Managed Identity This article demonstrates how to securely connect a .NET app to Azure DocumentDB—with MongoDB compatibility—using Managed Identity from Azure Active Directory, eliminating the need for passwords in connection strings by leveraging runtime token authentication.
  • Tools for Azure AD B2C migration now available Migration tooling including Just-in-Time (JIT) migration, High-Scale Compatibility (HSC) mode, and detailed guidance helps Azure AD B2C customers transition to the newer, feature-rich Microsoft Entra External ID platform with minimal disruption.
  • Microsoft Entra ID security updates: What organizations need to do now Microsoft Entra ID is tightening security by retiring Custom controls in favor of External MFA, enforcing consistent Conditional Access during credential registration, and requiring explicitly registered authentication methods for self‑service password reset starting July 6, 2026.
  • Getting Secrets Out of YAML: Implementing Azure Key Vault CSI Driver on AKS with Workload Identity Implementing Azure Key Vault CSI Driver with AKS Workload Identity eliminates secret leakage by authenticating pods directly to Key Vault, reducing exposure risk and simplifying rotation workflows.
  • AD – Account Lockout (Free Tool) The AD Account Lockout tool from Microsoft provides real-time visibility into which Domain Controller is responsible for user account lockouts, helping IT professionals quickly pinpoint and resolve issues without relying on guesswork.

🔌 Integration

  • Dynamic Connection Properties in Azure Logic Apps Standard Azure Logic Apps Standard now allows dynamic connection names for built-in connectors, enabling workflows to switch between multiple SFTP servers or databases at runtime using expressions, improving scalability without separate workflows.
  • Two Outlook Connectors in Copilot Studio; What’s the Difference? The two Outlook connectors in Copilot Studio differ primarily in purpose and functionality; Office 365 Outlook is a versatile, Power Platform‑compatible connector with extensive actions for automation across platforms, while Microsoft 365 Outlook is tailored for generative AI agents with curated tool groups to simplify natural language interactions.
  • Productize, observe, version, and automate MCP servers in Azure API Management Azure API Management now lets you package, govern, observe, version, and automate MCP servers as first‑class resources, enabling productized, observable, and automatable integration layers for agentic workflows.
  • Logic Apps Aviators Newsletter - June 2026 Logic Apps Aviators Newsletter - June 2026 features Florian De Langhe, Lead Expert/Team Lead at Delaware, who emphasizes the importance of documentation in integration projects and advocates for finer control over Logic Apps scaling to enhance workflow efficiency.
  • GA of NSP for Azure Service Bus & NSP now available in Azure Gov. Regions Network Security Perimeter's General Availability for Azure Service Bus in Azure Government regions—including Texas, Arizona, Virginia, DoD East, and DoD Central—enables customers to enforce perimeter‑based governance on messaging services while meeting strict compliance requirements.
  • Read profiles without chasing every red bar Start with memory profiling to identify allocation hotspots that compound over time, using filters to focus on your code's ownership boundaries and domain context before diving into CPU analysis for performance shaping insights.
  • Power Platform tenant inventory — a community showcase of the API-first management surface VerseOps is an open‑source Power Platform tenant inventory tool that provides detailed offline snapshots, diff capabilities between snapshots, joined views of capacity and asset counts, raw JSON inspection for each row, and a lightweight codebase designed to be forked and extended by administrators.
  • A message queue bought us time Introducing a sequential message pump protects downstream systems but can cause queue wait times to increase under load, while a fire‑and‑forget approach speeds up perceived performance until it leads to unbounded concurrency issues.

💡 Internet of Things

🎓 Learning and Certifications

⚖️ Management and Governance

  • How to Measure Token Impact of MCP Tool Invocation in Microsoft Foundry This guide explains why token counts from API usage, portal traces, and trajectory views can diverge in Microsoft Foundry’s MCP tool invocations, offering a reproducible method to reconcile these metrics for accurate FinOps reporting.
  • Right‑sizing Azure Savings Plans, one hour at a time Azure Savings Plans can now be fine‑tuned hour by hour using the new Benefit Recommendations API, which provides detailed PAYG usage data and alternative commitment levels to help FinOps teams optimize spend more accurately.
  • Modern VM monitoring, powered by OpenTelemetry OpenTelemetry Guest OS metrics in Azure VMs offer lower-cost, faster queries, per‑process visibility for deeper troubleshooting, consistent cross‑OS monitoring, native PromQL support, and open‑standardization across observability tools.
  • At-Scale Failure Reporting for Azure Update Manager At-Scale Failure Reporting for Azure Update Manager introduces a method using Azure Resource Graph to aggregate and query patch failure data across thousands of machines in a single operation, significantly improving visibility and troubleshooting efficiency.
  • What We Took Away from Build 2026 as Microsoft Field and Solution Engineers At Build 2026, four Microsoft Solution Engineers highlighted how tools like Scout automates scheduling and Agent 365 manages rapidly proliferating agents with identity governance, reflecting a shift from AI assistance to AI‑driven automation that addresses real enterprise challenges.
  • PUBLIC PREVIEW - Azure Monitor - Collect Azure Resource Platform Logs at Scale with DCRs Azure Monitor’s public preview of Data Collection Rules (DCRs) simplifies collecting platform logs across thousands of Azure resources by allowing a single rule to govern configuration, cost, security, and governance for diagnostics settings.
  • Agentic AI vs. Traditional Automation: When Should an Enterprise Use Agents? Agentic AI excels in complex, decision‑heavy tasks where traditional automation struggles with ambiguity and requires continuous learning, making it ideal for enterprises needing adaptive solutions.
  • How ARM Tracks Work That Takes Hours ARM's long-running operation protocol uses Azure-AsyncOperation and Location headers, Retry-After values, and provisioningState properties to let clients asynchronously track hours-long resource work through polling status URLs until completion.
  • Azure Availability Zone Mapping and VM Resilience Analysis Guidance using SRE.AZURE.COM Agent The SRE.Azure.com agent helps Azure engineers map logical Availability Zones to their physical counterparts, analyze VM distribution across zones, and identify resilience gaps—critical for ensuring high availability, disaster recovery, and compliance despite differing zone mappings across subscriptions.
  • Azure Bicep at Scale: Architecture, Repos, and Landing Zones Done Right Azure Bicep at Scale introduces best practices for organizing architecture repositories and establishing consistent landing zones to streamline large‑scale deployments.
  • How to Monitor Windows Servers with Prometheus and Grafana This guide shows how to use the windows_exporter alongside Prometheus and Grafana to monitor Windows servers securely, configure scraping for optimal performance, and set up Alertmanager for timely notification of issues.
  • Azure Subscription Types Azure Subscription Types explains how Management Groups enforce governance across multiple subscriptions, with Pay-As-You-Go offering instant deployment for startups while Enterprise Agreements provide custom commercial terms and centralized billing for large enterprises.

🎬 Media

  • Online Meeting Effectiveness Index (OMEI) The Online Meeting Effectiveness Index (OMEI) uses Azure AI to evaluate virtual meetings on Participation, Engagement, Structure, Sentiment, and Tech Quality, revealing that while tech quality scores high at 92/100, engagement lags at 77/100, suggesting clearer agendas and interaction points can

🚌 Migration

  • Public Preview: Azure Site Recovery support for Linux Azure VMs with NVMe disk controllers. Azure Site Recovery now enables replication and disaster recovery for Linux VMs equipped with NVMe disk controllers on Gen 2 families like Da/Ea/Fa v6‑series and Ebsv5/Ebdsv5, enhancing performance‑critical workloads’ resilience.
  • Why even the best teams struggle with cloud migrations Redgate’s 2026 report reveals that even experienced teams face significant cloud migration challenges primarily due to "unknown unknowns"—unforeseen issues arising from data integration complexities, system intricacies, skill gaps, and new security requirements.
  • Microsoft Options to Migrate SQL Server Databases The article outlines various migration tools—such as Azure DMS, Managed Instance Link, and SqlPackage—with details on supported targets, downtime requirements, and limitations for moving SQL Server databases to Azure SQL Database, MI, or VM environments.

🌐 Networking

🔐 Security

  • Public Preview: Microsoft Foundry agent security capabilities in Microsoft Defender for Cloud are transitioning to a Microsoft Agent 365 license The upcoming transition of Foundry agent security features from Defender for Cloud to the Agent 365 license will enhance threat detection with unified observability logs starting July 1, 2026.
  • DLP for Copilot Prompts: How to Stop Sensitive Data From Leaving Your Tenant Microsoft has introduced Data Loss Prevention controls for Copilot prompts, allowing IT teams to block or restrict web grounding based on sensitive information types or keyword lists, with options to simulate policies before deployment, all under the Purview Information Protection and Governance SKU.
  • Reduce Microsoft Sentinel Ingestion Costs with Smarter Tiering Reduce Microsoft Sentinel ingestion costs by applying smart tiering strategies that filter high-volume, low-impact logs and direct them to less expensive storage options, validated through KQL analytics.
  • DoH is now generally available on Windows DNS Server DoH’s general availability on Windows DNS Server now allows organizations to securely encrypt and authenticate client-to-resolver DNS traffic directly within existing on‑premises infrastructure, enhancing privacy, reducing spoofing risk, and supporting Zero Trust networking without overhauling current resolver setups.
  • Reduce unnecessary internet exposure with Microsoft Defender Microsoft Defender introduces a new security recommendation that helps organizations identify and reduce unnecessary internet exposure on devices accessible from the public internet, providing context to understand why exposure exists and guiding actions to secure high-risk assets.
  • Participatory Extensible Security Participatory Extensible Security (PES) enables clients to request security assessments from any provider via APIs like AMSI, improving flexibility and performance while risking crashes and silent performance degradation if providers are unreliable or unsigned.
  • Introducing scheduled antivirus scans on Microsoft Defender Linux Microsoft Defender now offers centrally managed scheduled antivirus scans for Linux, allowing security teams to configure hourly quick scans, daily scans at defined times, or weekly scans with flexible options like idle detection, low CPU priority, and definition updates before scanning.
  • Reconstructing AI activity in investigations The new investigator playbook from Microsoft provides a structured method for reconstructing AI activity in investigations involving Copilot and Azure AI services, turning telemetry into coherent accounts of events across identity, timing, resources, and detection signals.
  • Agent 365 | Security Operations in Defender Agent 365 integrates security operations across Defender, the M365 admin center, and Entra to enable real-time blocking of malicious tool calls, detection of jailbreak attempts on Copilot Studio agents, and automated revocation of compromised sessions.
  • Elevate your telemetry using custom data collection in Microsoft Defender Custom data collection in Microsoft Defender lets security teams capture specialized telemetry—such as AMSI for script content and Kerberos for authentication attacks—directly within the portal, enhancing visibility into command and control communications with dynamic tagging for targeted high‑value endpoints.
  • Microsoft Defender now monitors RPC activity Microsoft Defender now monitors specific RPC operations associated with common attack techniques like lateral movement, credential theft, and privilege escalation, providing granular telemetry to detect and disrupt malicious activity without requiring visibility into the source device.
  • AI brands as bait: How threat actors are using the AI hype in social engineering Threat actors are using AI brand names like ChatGPT and Claude in phishing campaigns that collect credit card data and credentials, highlighting a new wave of social engineering attacks leveraging artificial intelligence hype.
  • The Worm in the Supply Chain: How Defender for Endpoint and Sentinel for SAP BTP Caught Shai-Hulud The attack, dubbed "Mini Shai-Hulud," targeted four critical SAP npm packages—@cap-js/sqlite, @cap-js/postgres, @cap-js/db-service, and mbt—used in CAP development and CI/CD pipelines, highlighting how even enterprise-focused tools can become supply

📦 Storage

💻 Virtual Desktop Infrastructure

  • MDOP is out of support: What to do next with Microsoft Intune MDOP's end of extended support on April 14, 2026, necessitates migrating to supported alternatives like Intune, Windows 11, and Entra ID, which offer unified cloud management and enhanced security within a Zero Trust framework.

🔗 Web

  • Now in preview: built-in MCP for Azure App Service Built-in MCP for Azure App Service automatically converts hosted REST APIs into Model Context Protocol servers, enabling AI agents to call them with minimal configuration and handling protocol negotiation, tool discovery, and spec hot reloading.
  • What's new in Azure App Service at #MSBuild 2026 Azure App Service’s #MSBuild 2026 launch introduces an Easy AI experience that lets web apps become AI‑ready instantly, along with Isolated v4 for enhanced security and compliance, Managed Instance for seamless legacy migration, and platform CLI upgrades to streamline deployments—accelerating AI application development
  • Debug App Startup Faster on Azure App Service for Linux with Startup Logs The new Azure CLI commands for Azure App Service for Linux provide detailed startup logs, automatically prioritizing failure logs and including hints for common issues like port mismatches and container timeouts, streamlining debugging of application startup problems.
  • ASP.NET Core background tasks with NCronJob and SignalR NCronJob integrates with ASP.NET Core SignalR to send real‑time, scheduled messages via Razor Pages UI, demonstrating both concurrent and non‑concurrent job execution with a simple Hub implementation.

Get Azure Weekly in your inbox every Sunday

Sign up to receive the email every week and keep on top of all the announcements.

Don't forget to confirm your subscription. We have a double opt-in sign up process to prevent spam.

We will never provide your details to any third parties. We don't spam.

Azure Weekly Logo