Microsoft has introduced Azure Catalyst, an agent-led service intended to help organizations discover, plan and execute modernization across applications, data and infrastructure. Announced on November 17, the offering is part of a broader set of Azure updates focused on building AI applications and running them across cloud and distributed environments.
Microsoft says Azure Catalyst uses AI agents to analyze legacy code, identify dependencies and produce modernization plans, then assist with execution. The company positions the service as a way to address technical debt and constrained developer capacity, with work that traditionally takes months or years potentially completed in weeks or days. Those timing comparisons are Microsoft’s characterization; the announcement does not provide a general benchmark or methodology.
A connected path from discovery to execution
Azure Catalyst is designed to connect IT teams and developers through an end-to-end workflow. New capabilities described by Microsoft include estate insights, AI-powered application discovery, modernization strategy and cost comparisons, and AI-generated migration waves. The execution scope spans applications, containers, storage, networking and databases.
The service builds on existing support for Java and .NET modernization and migration assistance for VMware, Hyper-V and physical servers. Microsoft also announced preview capabilities for modernizing applications and databases together, as well as migrating from BizTalk to Azure Logic Apps. These capabilities are relevant to organizations planning broader cloud transitions, where application dependencies and data platforms can make isolated infrastructure moves difficult to sequence.
Microsoft says customers can apply governance rulebooks and bring their own models to guide agent behavior. Reusable runbooks and skills are intended to make successful procedures repeatable across projects. The controls are meant to let organizations align proposed plans with their architecture, priorities and security requirements rather than relying on a single preset approach.
Security and AI readiness in the modernization workflow
The announcement links modernization with vulnerability remediation. Microsoft says Azure Catalyst works with Microsoft Security to scan and rank risks, prioritize actions according to severity and exposure, and identify vulnerable code and databases. Teams can use GitHub Copilot to help remediate code as applications are modernized, according to the company.
Azure Catalyst also works in GitHub Copilot to help teams understand business logic and dependencies in complex applications and upgrade them to Azure managed services. Microsoft’s stated aim is to make these systems easier to connect to AI workflows after modernization. That approach makes application understanding and migration planning part of the same process, rather than treating the cloud move as a separate infrastructure task.
Microsoft reported that 1.4 million applications have been modernized using the service and cited Levi Strauss as a customer that modernized enterprise systems and deployed more than 1,000 agents across them. The announcement also describes Microsoft’s internal use of the tooling: Microsoft Quantum and Xbox achieved 80% faster application modernization, while Intune upgraded a key library 75% faster than anticipated. GitHub ported an 800,000-line production-code agentic harness from TypeScript to Rust. These are company-reported results and do not establish expected outcomes for other organizations.
Modernization alongside Azure’s AI infrastructure work
Microsoft presented Azure Catalyst alongside updates intended to support the growing infrastructure requirements of AI workloads. For Kubernetes environments, the company announced an Agent Sandbox extension for Azure Kubernetes Service (AKS), which runs agent-generated code in managed sandboxes with virtual-machine-level isolation for each session. Microsoft also described pre-warmed pools and a serverless option. Agent Grid is intended to provide authenticated, encrypted connections among agents, models, tools, data and Azure services, with identity-based policies and default-deny network isolation.
Other updates address context reuse and workload placement. Microsoft says long-running agents can reuse processed prompt context rather than recomputing it on GPUs, with cached context stored durably in a customer’s Azure tenant. AKS AI Manager is designed to match deployments to GPU capacity and other infrastructure conditions across the fleet. The company also highlighted a connected path between Microsoft Foundry and AKS, including support for moving open-weight models and agents from development into production environments.
For organizations modernizing estates while preparing for AI, the announcements connect application transformation with choices about runtime, infrastructure and governance. Azure Catalyst’s announced coverage includes application and database work as well as server and virtualization migrations; the surrounding Azure capabilities are intended to extend AI workloads across cloud and distributed locations. Microsoft has not specified in this announcement a single availability date for every new capability, and several are described as previews.