Agents now sit at the center of modern applications, calling models, tools and other agents across a stack of APIs, data and infrastructure, so when a customer says the agent is slow, the cause can sit anywhere. At Microsoft Ignite 2026, Azure Monitor gives developers, SREs and IT operations teams one platform for end-to-end visibility across agents, applications, data and infrastructure. This year's announcements span four areas: full-stack agent observability, AI-assisted operations, business impact with Microsoft Fabric, and cost efficiency at scale.
Why this matters
Agentic applications are changing what it means to keep a service healthy. An agent might call several models and tools, hand work off to other agents, and rely on APIs, databases, Kubernetes clusters and virtual machines behind the scenes, so when a customer says the agent is slow, the cause can sit anywhere in that stack. At the same time, technical health needs business context to be meaningful, and both model consumption and telemetry add cost at scale.
At Microsoft Ignite 2026, Azure Monitor is responding with one platform that gives developers, SREs and IT operations teams end-to-end visibility across agents, applications, data and infrastructure. This year's announcements fall into four areas: full-stack agent observability, AI-assisted operations, business impact with Microsoft Fabric, and cost efficiency at scale.
Full-stack agent observability
Agent observability follows the same loop your teams use to improve agents: observe what happened, evaluate quality, then optimize. Azure Monitor and Microsoft Foundry share one platform and the Microsoft OpenTelemetry Distro, so builders and operators work from the same telemetry.
Observe and monitor: see what happened
Agent details in Application Insights show per-agent metrics, errors, tool calls, models and tokens, correlated with application and infrastructure signals so you can pinpoint where failures begin. You can also link and observe agents built outside Foundry (generally available), compare health and usage across all your agents in a new fleet view (public preview), and track token usage and cost per agent (private preview). Learn more: aka.ms/AzMon_AgentObs | Unified Agent observability with end-end tracing in Azure Monitor
Trace Analytics builds on Application Insights distributed tracing to accelerate root cause analysis at-scale. It stitches spans across resources into key transactions, drives metrics for SLIs, alerts and dashboards, and lets you compare healthy and failing transactions at scale, including GenAI interactions. Learn more: aka.ms/AzMonTracesPreview
Evaluate: measure quality
A new evaluations view brings Foundry evaluations into Azure Monitor so you can track agent quality at scale. In Foundry, annotations add human feedback to agent runs, and a safety evaluator scores responses against safety rubrics. Learn more: Monitor AI agents in Application Insights and Run evaluations in Foundry.
Microsoft Foundry turns those signals into action, with skill optimization from evaluation results, suggested fixes for failing runs and cost optimization to reduce token and model. Suggested code changes for agents are coming soon. Learn more: [Link to Foundry agent optimization blog]
A secure, private, multimodal platform
Protected tables (public preview) give GenAI content such as prompts and responses its own RBAC, and PII redaction (public preview) detects and masks values such as emails and SSNs in Log Analytics. Cross-agent debugging (public preview) brings in telemetry from Copilot Studio agents, GitHub Copilot and Azure Copilot, with Fabric Data Agent, Microsoft 365 Agent Builder, Microsoft 365 Copilot, Cowork and OTLP agents coming soon. You can also inspect images, PDFs and audio in agent interactions (private preview), and a lower-cost SKU for agent telemetry with a starter kit and free tier is coming soon. Learn more: aka.ms/ProtectGenAIContent | aka.ms/ProtectGenAIPii
Find and close monitoring gaps across the stack
OpenTelemetry onboarding is now generally available for apps on Azure VMs, scale sets and AKS, and AKS control plane metrics are generally available. Monitoring Coverage gives you one place to see coverage for VMs, AKS and Arc-enabled Kubernetes and apply recommendations in batches of up to 100 resources. New platform telemetry coverage (preview) extends this to resources such as storage accounts and key vaults, with data flow validation that flags problems before they cause data loss. Learn more: aka.ms/AzMonOTLPIgnite2026 | Monitoring coverage in Azure Monitor - Azure Monitor | Microsoft Learn
AI-assisted operations
Bring Azure Monitor into the coding agents your teams already use, and track workload health with flexible health models and SLOs.
Azure Observability skills plugin in coding agents (public preview)
The new skills plugin brings Azure Monitor capabilities into coding agents such as GitHub Copilot in VS Code, so you can detect, investigate, remediate and validate issues end-to-end. The skills draw on Prometheus and the Log Analytics, Application Insights, traces and Foundry MCP servers, packaging best practices so tasks use fewer tokens. A minimal fix is proposed for human approval, then evaluations confirm recovery before anything reaches production. Learn more: aka.ms/DashboardWithGrafanaMCP | aka.ms/AzObsSkillsPlugin
Azure Monitor health models (generally available)
Health models turn monitoring signals and dependencies into a clear view of workload health that both teams and agents can use. Start with SLIs and SLOs for a Service Group or any set of resources, then nest component models into a workload view that shows dependency impact and blast radius. Health is also available in Resiliency Hub and through a Health API and MCP server. Learn More: http://aka.ms/ahmgablog
Business impact with Fabric
Azure Monitor logs mirroring to Fabric (public preview contd.)
Mirroring Azure Monitor logs to Microsoft Fabric lets you correlate errors and latency with transaction and customer data to assess revenue at risk and prioritize incidents by business impact. Service health and business KPIs then come together in Power BI as a shared view for IT and business stakeholders.
Cost efficiency at scale
More data means more cost, so Azure Monitor gives you two levers: match storage to how data is used and reduce telemetry before it is ingested.
Match log storage and retention to your needs
Analytics Logs serve real-time monitoring and alerting, while Basic and Auxiliary Logs offer pay-per-query pricing for troubleshooting and high-volume audit data. Long-term retention keeps data for up to 12 years, and commitment tiers optimize pricing
Azure Monitor pipeline (public preview additions)
The pipeline adds Azure Monitor's first managed OpenTelemetry collection for logs, metrics and traces at terabyte scale, plus Linux VM and server hosting so you can run it without Kubernetes. It filters and compresses data with KQL before ingestion and routes to Azure Monitor or any OpenTelemetry destination, with generally available health signals that flag potential data. Learn more: aka.ms/AzMonVMSupport
Shape data at ingestion
Enhanced data transformations (public preview) give you more control over what lands in your workspace. Schemaless ingestion keeps undeclared properties in a dynamic column, so evolving logs no longer need a schema update for every field, and a new data collection rules overview helps you spot duplicate collection that adds cost. Learn more: aka.ms/SchemalessIngestion
Get started
Customers are already seeing what that full-stack visibility makes possible. As Chris Waugh, Manager of Platform Architecture at Radiology Partners, puts it: “Azure Monitor is the cornerstone of Radiology Partners’ cloud operations strategy, giving us centralized visibility across critical Azure services, including virtual machines, AKS, Application Gateways, and more. With Log Analytics, Application Insights, and real-time alerting, our teams have the insights they need to prevent outages and quickly diagnose issues when they occur. This has helped us prevent downtime and, when issues do arise, gives us a jumpstart understanding of what the issue may be. Tailored alerts through Microsoft Teams bring the right context to the right teams, helping us resolve issues faster and keep our cloud environment reliable and resilient.” – Chris Waugh, Manager of Platform Architecture, Radiology Partners
Whether you are building your first agent or running hundreds of services in production, Azure Monitor gives you one place to observe the full stack, investigate with AI and human oversight, connect reliability to business outcomes, and keep costs under control. Try the new capabilities in the Azure portal and tell us what you think.