Browse All Posts (306)
Muzz Imam and Eamon O'Reilly outline hosted-agent updates announced at Ignite 2026, including network egress controls, private certificate trust, resilient execution, memory snapshots, and A2A passthrough. They explain availability, operational limitations, and the planned option to run Foundry's agent runtime on customer-operated Azure Kubernetes Service clusters.
Nandini Muralidharan and Abhinav Premsekhar introduce the generally available Browser Automation tool in Foundry Agent Service. They explain action batching, human intervention, and session recording, alongside preview capabilities for private networking, proxy routing, and secure sign-in that help developers build browser-based agent workflows.
Rong Lu introduces the generally available Foundry Dev Pack, a single install for Microsoft Foundry agent development tools. The article explains editor, terminal, and coding-agent workflows, including local debugging, GitHub Copilot project generation, repeatable Azure deployments, and setup commands for Windows, macOS, and Linux.
Jeff Hollan outlines Foundry Agent Service updates announced at Ignite 2026, including managed agent authoring, generally available voice and browser capabilities, and expanded hosted runtime controls. He also covers production tracing, evaluation, and optimization features, plus planned deployment to customer-owned Azure Kubernetes Service clusters.
Linda Li and Nandini Muralidharan explain how Microsoft Foundry Toolboxes share versioned tools, skills, and specialist agents through MCP. The announcement covers GitHub Copilot setup, monitoring and governance, expanded grounding and analytics connections, and generally available browser automation and computer-use tools, with Python examples for building a shared troubleshooting toolbox.
Joe Braley announces the November 2026 Microsoft Build of OpenJDK critical patch update, including fixes for Java4Shell in JDK 17, 21, and 25. The post identifies patched versions, explains why temporary Azure Web Application Firewall mitigations do not replace upgrading, and flags upcoming support deadlines.
Allison reports that GitHub security risk assessments now run automatically for organizations on GitHub Team. The free assessments help organization admins and security managers identify exposed credentials and code vulnerabilities, review their organization’s current risk, and prioritize next steps.
Leandro Iwase and Ehud Itshaki explain how Microsoft Entra Agent ID secures AI agents without giving them unrestricted access. The session covers separate agent identities, discovery, dynamic risk signals, and Conditional Access, with demonstrations of policy enforcement, malicious prompt blocking, and data loss prevention protections.
John Nix and Henrique Oliveira explain how Codename MDASH uses specialized AI agents and security-focused models to discover, validate, and prioritize code vulnerabilities. The session covers scanning architecture, chained exploits, security analyst workflows, and developer remediation through repository findings, pull requests, and validation of fixes.
Microsoft Source details Microsoft and Intel’s expanded Azure infrastructure collaboration, covering advanced silicon packaging, CXL memory, accelerated SQL backups and confidential computing. The announcement includes benchmark results for M-series v4 virtual machines and describes encryption of memory and CPU register state when confidential VMs move between hosts.
Mark Russinovich and Pavan Davuluri discuss infrastructure for running production AI across Azure, datacenters, edge environments, and Windows devices. The session covers AI infrastructure primitives, AKS and Foundry workloads, a factory inference demo, application modernization with Azure Catalyst, and security considerations for cloud and local AI.
Jason Moore and Alice Appleton review recent Copilot Studio changes and upcoming capabilities at Microsoft Ignite 2026. The session covers a unified maker experience for agents, workflows, and apps, natural language agent creation, improved agent performance, memory, model selection, and tools for governance, evaluation, and cost control.
Julia Kasper explains how the GitHub Copilot team evaluates and optimizes its coding agent across models, clients, and configurations. This GitHub Universe 2026 session covers reproducible benchmarks, autonomous coding runs, LLM-graded assertions, internal testing, and online experiments used to detect regressions and assess quality.
Hong Gao explains how Azure Monitor connects telemetry across managed and external AI agents using OpenTelemetry. The article outlines tracing support and shows how developers and site reliability engineers can follow agent handoffs, investigate model costs, and locate failing dependencies, provided telemetry collection and trace context propagation are enabled.
dmalbrough outlines the general availability of Azure Native Commvault Cloud, which brings Azure VM backup and recovery into the Azure portal. The service supports policy-based protection, rule-based VM discovery, flexible recovery options, and immutable backups, with managed identity access and applicable Commvault charges included on the Azure bill.
The Visual Studio Code Team outlines changes in the 1.148 Insiders build, including clearer agent session categories, configurable Inbox Mode navigation, relative times for scheduled messages, and revised editor tab behavior. These release notes track incremental improvements to everyday editor and agent workflows.
Dave Salvator and Maraki Ketema explore NVIDIA accelerated computing on Azure for AI, data science, and simulation workloads. The session covers infrastructure efficiency, confidential computing, Azure Local, and GPU-accelerated Microsoft Fabric Data Warehouse, including an architecture and performance demonstration and approaches for moving workloads from experimentation to production.
Kiran_Cherukuwada introduces Azure Elastic SAN Next Generation in public preview, covering higher per-volume performance, Direct Attach connectivity, and support for 20,000 volumes per SAN. The announcement explains capacity provisioning, shared performance, snapshot recovery, and regional availability, with a workload-specific cost comparison and guidance for evaluating preview limitations.
Brendan Powers and Srikumar Nair demonstrate a repeatable approach to assessing AI agent risk with Microsoft Agent 365. The session covers Shadow AI discovery, device and user context, blocking and isolation controls, and runtime enforcement, with Zenity’s perspective on investigating agent behavior and moving from discovery to governance.
Guhan Sivaji demonstrates how Reltio supplies trusted enterprise data to Microsoft Fabric and Foundry. The session covers duplicate-record investigation, governed data sharing, analytics and semantic models, and an agent workflow that uses a Reltio MCP server to access master data context.
Maya Nielan compares options for building Claude-powered agents on Microsoft Foundry, focusing on control versus managed hosting. The session separates model, harness, and infrastructure choices, explores Microsoft and Anthropic platforms, and explains how MCP and skills can keep tools portable across agent stacks.
Ajay Mungara explains how agentic AI workflows can combine local models with Azure capabilities. The session examines workload routing based on privacy, cost, and model capability, including escalation from smaller models to more sophisticated reasoning, alongside enterprise security and confidential computing considerations.
Zachary Symm explores identity-first recovery across Microsoft Entra ID, Active Directory, Microsoft 365 data, and Azure infrastructure. The session covers recovery sequencing, immutable backups, air-gapped protection, and tested runbooks, with attention to rebuilding clean environments and prioritizing the workloads needed to restore critical operations.
Microsoft Events explores how OneLake and Fabric IQ bring distributed enterprise data into an AI-ready foundation. The session covers catalog federation, Apache Iceberg interoperability, semantic models, and ontologies, with demonstrations of cross-cloud querying, zero-ETL integration, sensitive-data classification, and governance.
Andy Tran explains how to build a search pipeline that retrieves company data to ground AI responses. The session covers Cohere parsing, embedding, and re-ranking models, with demonstrations of document extraction, multimodal retrieval, and product searches that must satisfy multiple conditions.
EY Kalman examines how to secure and scale enterprise Model Context Protocol servers. This advanced, demo-first session covers stateless deployment, multi-tenant routing, authentication, prompt-injection mitigation, and operational governance, drawing on FastTrack deployments and an Azure reference architecture to address production risks and protect corporate data boundaries.
Alessandro Cesarini and Nicolas Groh demonstrate how to investigate and recover a broken access path across Microsoft Entra ID and Intune. The session covers tracing configuration changes, comparing policies against a known-good state, and restoring only what is needed, alongside guardrails and recovery preparedness.
Matt Egen demonstrates how organizations can use the security capabilities already included in Microsoft 365 E5. The session covers attack disruption, identity and data protection, conditional access, and integrated security operations, with a live demo of workspace setup, connectors, UEBA, and threat intelligence.
Scott Manchester explores paths for modernizing desktop delivery with AVD Hybrid and Windows Cloud. The session covers migration planning, demonstrations of Nerdio Compass assessments for Microsoft and Citrix environments, ongoing optimization with Nerdio Visor, and workflows involving vSphere, Nutanix, and combined on-premises and cloud resources.
Christina Mallon demonstrates how inclusive design can improve AI-powered experiences across the product lifecycle. The session covers identifying accessibility gaps, researching with diverse participants, designing for permanent and situational disabilities, and evaluating AI inclusivity through disability representation datasets and practical accessibility testing.
Tiffany Songvilay explains how to govern Copilot agents according to their risk, data access, and ability to take actions. This session covers lifecycle oversight, ownership, environment strategy, and sensitive data protection using Microsoft Purview, Copilot Studio, and DLP controls, with guidance for assessing production readiness and preventing agent sprawl.
Maddy Montaquila introduces Aspire 17, covering shared team configuration, a redesigned dashboard, preview Go AppHosts, and stable .NET and Kubernetes integrations. The release also updates coding-agent guidance and debugging workflows, with migration considerations for .NET requirements, project builds, secrets storage, and Azure Kubernetes Service control-plane costs.
Dave Byrne, Rajesh Setlem, and Sandeep Lingam explore the architecture and internals of Azure HorizonDB, a cloud-native PostgreSQL service. The session covers its storage engine, separation of compute and storage, replication pipeline, recovery design, and elasticity mechanisms for predictable performance and resilience at scale.
Mark Sentell, Sachin Patney, and Ben Zulauf demonstrate building enterprise applications with GitHub Copilot and Fabric data. The session covers analytical dashboards, AI-generated summaries, transactional planning tools, backend connectivity, authentication, and deployment, alongside governance controls and a Craneware Group case study.
Aditya Sharad and John Lam explain how GitHub Intelligence uses code, pull requests, reviews, and AI agent activity to identify recurring problems across teams. The session covers finding affected repositories, building remediation plans, and dispatching coding agents under human review, with results from Microsoft and customers.
Tina Schuchman, Hiroaki Kimura, and Jeff Hollan cover the agent development lifecycle in Microsoft Foundry, from local development with GitHub and VS Code to hosted deployment. The session explores governed tools, enterprise data grounding, tracing, evaluation, and cost insights before publishing agents to Microsoft 365 Copilot and Teams.
Carl Cortright explores how Replit builds and deploys on Azure, covering agent orchestration, cellular architecture, durable execution, and micro VM sandboxes. The session examines delegated permissions, enterprise data security, and deployment infrastructure, alongside Replit's reported threefold increase in per-engineer output while keeping incidents flat.
Yitzhak Kesselman, Adi Regev, Pankaj Arora, and Tessa Kloster explore how Microsoft Fabric supports mission-critical analytics and AI workloads. The session covers resilience, safe change rollouts, and data protection, alongside practical guidance for capacity, governance, and deployment as usage grows, with examples from customers.
Rahul Bagaria and Evgeny Ternovsky explore observability across infrastructure, applications, and AI agents with Azure Monitor. The session covers health models, platform metrics, and Observability Agent capabilities for investigating issues, alongside Microsoft Fabric’s role in connecting operational and business data.
Silvano Coriani introduces Microsoft Fabric HorizonDB, a Fabric-native PostgreSQL experience for building applications that combine transactions, analytics, and AI. The session covers querying operational and analytical data through one PostgreSQL experience, reducing data movement and ETL complexity, and accessing OneLake-scale analytics with built-in governance.