Azure workflows for coding agents that help you prepare deployments, investigate cloud issues, and manage resources.
Azure Skills is Microsoft’s marketplace of plugins for working with Azure from a coding agent. Its workflows help developers and platform engineers choose services, prepare applications, investigate running systems, and manage cloud costs. The plugins combine task guidance with MCP connections for working with Azure resources.
Describe your task and the agent can select relevant skills. In Claude Code and GitHub Copilot CLI, you can also select a skill explicitly; Gemini CLI asks for consent when it activates one. Installed plugin hooks run automatically on their configured events. Which skills, hooks, and MCP connections you receive depends on the plugin and installation method.
Assess an existing codebase for deployment blockers, choose Azure services, and prepare a deployment plan. For projects using Azure Developer CLI, the preparation, validation, and deployment skills cover successive parts of that workflow; deployment expects a prepared and validated project.
With the optional aks-skills plugin installed, collect read-only evidence for crashing pods, unhealthy nodes, or connectivity failures. Its skills distinguish general investigations, exact known-error matches, and GPU inference incidents. Packet capture is a separate workflow for explicit requests after initial diagnostics and requires approval.
Install the optional azure-cost plugin to break down actual spending and investigate changes, including AKS and AI costs. Separate workflows handle forecasts, pricing estimates, resource savings, and budgets so a billing investigation stays distinct from a recommendation or planning estimate.
Check Azure Functions and App Service deployments for zone redundancy, storage replication, health probes, and regional failover. The reliability skill can guide confirmed remediation through CLI changes or infrastructure code; its assessment does not cover Container Apps.
Install the optional foundry-iq-skills plugin to make supported local files, Blob Storage, or Data Lake documents searchable, retrieve answers with citations, and connect a knowledge base to an existing agent. Its helper scripts require Python 3.12 or later; the plugin documents Copilot CLI support on Windows and Linux.
With the optional azure-kusto-graph-skills plugin installed, translate a threat-hunting question into an IRQL pipeline, build native Kusto graphs for relationship analysis, or visualize existing query results with IRQL graph functions. These are separate skills for composing queries, analyzing graphs, and presenting results.
Install the optional azure-local-skills plugin to plan deployments, investigate failures, and manage workloads with procedures matched to your Azure Local deployment scale. Standard deployments and multi-rack deployments use separate skills; multi-rack is in preview, and the procedures are not interchangeable.
The catalog spans the base azure plugin and six optional specialist plugins. Specialist skill descriptions name the plugin to install; the other skills belong to the base plugin. Install coverage varies by client; Claude Code and Copilot CLI have specialist installation commands under Install. The hook rows represent separate client configurations and package locations; they are not a sequence of hooks to configure manually.
Plugin hooks load with their enabled plugin and report session metadata or matching usage events. Their telemetry helper downloads a pinned native reporter when needed, verifies its digest, and caches it. To stop this telemetry while keeping the skills, set the agent process’s AZURE_MCP_COLLECT_TELEMETRY environment variable to false before launch.
Showing all 46
Check code, dependencies, and framework compatibility before planning infrastructure.
Plan and deploy an application with estimates and pre-deployment approval.
Generate deployment files for projects explicitly using Azure Developer CLI.
Check deployment configuration, infrastructure, prerequisites, and identity permissions.
Execute deployments after the project’s preparation and validation are complete.
Deploy Flask, Django, or FastAPI code to Linux App Service.
Assess cross-cloud workloads and convert supported applications for Azure.
Assess Azure service migrations, hosting upgrades, and legacy Java SDK modernization.
Design enterprise infrastructure and generate Bicep or Terraform without azd.
Choose and provision VMs, scale sets, and capacity reservations.
Choose storage services and tiers, manage data, and plan lifecycle policies.
Configure Entra app registrations, OAuth authentication, and MSAL integration.
Plan AKS clusters, configure operations, and assess workload resource sizing.
Set up AI Runway on AKS through the first model deployment.
aks-skills: Investigate live AKS incidents through read-only evidence collection.
aks-skills: Match complete AKS error signatures to documented causes and fixes.
aks-skills: Investigate running AKS GPU and KAITO workloads with read-only evidence.
aks-skills: Collect approved, bounded packet captures after initial AKS diagnostics.
aks-skills: Analyze AKS placement, scaling bottlenecks, NAP readiness, and storage efficiency.
aks-skills: Find portal blades for supported Kubernetes and Compute resources.
Find and inventory Azure resources across subscriptions, groups, and tags.
Turn resource-group relationships into Mermaid architecture diagrams.
Investigate Azure production failures using logs, health signals, and supplied evidence.
Guide Application Insights instrumentation, SDK setup, and application telemetry patterns.
Troubleshoot Event Hubs and Service Bus SDK connections and message handling.
Assess and improve resilience for Azure Functions and App Service.
Check regional capacity and usage, and manage Azure quota requests.
Audit Azure security posture, compliance, and Key Vault expiration risks.
azure-advisor: Review Advisor recommendations, service retirements, and recommendation status.
azure-advisor: Develop remediation guidance for Azure Advisor recommendations.
Find relevant catalog skills and their plugin installation instructions.
Work with AI Search, speech, OpenAI, and document extraction.
Build, deploy, evaluate, and manage Foundry agents and models.
foundry-iq-skills: Build and query knowledge bases, then connect them to existing agents.
Configure API Management policies for AI models, tools, and agents.
Provision Entra agent identities and configure their OAuth token exchanges.
azure-cost: Break down actual spending and investigate Azure bill changes.
azure-cost: Forecast spending and estimate prices for planned resources and workloads.
azure-cost: Find resource savings and analyze Reservations or Savings Plans.
azure-cost: Manage budgets, alerts, tagging checks, and cost-related policy restrictions.
Query logs, telemetry, and time-series data with KQL in Azure Data Explorer.
azure-kusto-graph-skills: Build Kusto graphs and analyze patterns, paths, and connected components.
azure-kusto-graph-skills: Translate security investigation questions into composable IRQL query pipelines.
azure-kusto-graph-skills: Map existing query results into enriched graphs for Kusto Explorer.
azure-local-skills: Plan and operate standard Azure Local deployments and their workloads.
azure-local-skills: Manage preview multi-rack deployments using NetworkCloud and managed network fabric.
No matching results.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs after MCP execution to report matching calls to the Azure server.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill and reference-file activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill and reference-file activity.
Runs after MCP execution to report matching calls to the Azure server.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill and reference-file activity.
Runs after MCP execution to report matching calls to the Azure server.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill and reference-file activity.
Runs after MCP execution to report matching calls to the Azure server.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill and reference-file activity.
Runs after MCP execution to report matching calls to the Azure server.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill and reference-file activity.
Runs after MCP execution to report matching calls to the Azure server.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill and reference-file activity.
Runs after MCP execution to report matching calls to the Azure server.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill, reference, or Azure MCP activity.
Runs at session start or resume to report plugin and client metadata.
Runs after tool use to report matching skill and reference-file activity.
Runs after MCP execution to report matching calls to the Azure server.
Connect the agent to Azure tools for resource discovery, diagnostics, and management.
Provide Azure Cost workflows with spending queries, pricing, and Cost Management operations.
Choose your tool to install the base Azure plugin or its documented extension equivalent. Claude Code and Copilot CLI also offer optional specialist plugin installs below; adding the marketplace alone does not install them. Plugin telemetry hooks can be opted out of by setting the agent process’s AZURE_MCP_COLLECT_TELEMETRY environment variable to false before launch.
Install the base azure plugin from Anthropic’s official marketplace, then optionally add specialist plugins from Microsoft’s Azure Skills marketplace. Azure MCP needs Node.js 22 or later with npx on PATH; live Azure work needs an identity with access to the selected resources.
Azure MCP downloads and runs through npx, which inherits your local .npmrc configuration. Install only on a trusted device with a trusted npm configuration: a compromised configuration could cause malicious code to run. Plugin hooks collect telemetry unless you apply the opt-out above the route chooser.
In Claude Code, open the base Azure plugin’s installation panel and choose its installation scope. If azure is already installed from either marketplace, keep that copy and skip this install.
/plugin install azure@claude-plugins-officialIf you want specialist workflows, add Microsoft’s Azure Skills marketplace in Claude Code. Keep exactly one copy of azure: adding this marketplace for specialists does not require installing azure again.
In Claude Code, add the marketplace once, even if your base plugin came from Anthropic’s official marketplace.
/plugin marketplace add microsoft/azure-skillsThen run only the commands for the specialists you want. Each command opens that plugin’s installation panel; choose its scope there.
aks-skills: AKS incident diagnosis, GPU inference, packet capture, and cluster efficiency. Run in Claude Code.
/plugin install aks-skills@azure-skillsazure-cost: Spending analysis, estimates, optimization, and governance. Run in Claude Code.
/plugin install azure-cost@azure-skillsazure-advisor: Advisor recommendations and remediation guidance. Run in Claude Code.
/plugin install azure-advisor@azure-skillsazure-kusto-graph-skills: Kusto graphs, IRQL investigations, and graph visualization. Run in Claude Code.
/plugin install azure-kusto-graph-skills@azure-skillsazure-local-skills: Standard and multi-rack Azure Local workflows. Run in Claude Code.
/plugin install azure-local-skills@azure-skillsfoundry-iq-skills: Knowledge-base workflows. Helper scripts need Python 3.12 or later; Microsoft documents its task support in Copilot CLI on Windows and Linux. Run this installation command in Claude Code.
/plugin install foundry-iq-skills@azure-skillsFor Azure China or US Government, configure the MCP server and CLI identity for that cloud before starting the server or checking access: sovereign-cloud setup.
If plugin changes remain pending, run this in Claude Code to activate them before checking availability. Forced reload costs one uncached request; restarting Claude Code is an alternative.
/reload-plugins --forceOpen the plugin manager’s Installed tab and confirm that azure and each specialist you selected are enabled. Type a slash in chat and look for skills under their plugin names. If you installed azure-cost, also inspect the MCP connections for its Azure Resource Manager server.
/pluginFor the Azure connection check, authenticate Azure CLI, then ask “List my Azure resource groups.” Confirm that the tool-backed response uses your intended account.
Authenticate Azure Developer CLI separately when using azd deployment workflows.
Microsoft’s VS Code route installs the Azure MCP extension and its companion Azure skills extension for the base Azure workflows. It does not separately install the marketplace’s specialist plugins. Have Git installed for the skills extension and Node.js 22 or later with npx available for Azure MCP.
Azure MCP downloads and runs through npx, which inherits your local .npmrc configuration. Install only on a trusted device with a trusted npm configuration: a compromised configuration could cause malicious code to run.
Run this in your terminal to install the Azure MCP extension in VS Code.
code --install-extension ms-azuretools.vscode-azure-mcp-serverFor Azure China or US Government, configure the MCP server and CLI identity for that cloud before starting the server or checking access: sovereign-cloud setup.
Reload or restart VS Code so it can index the installed skills and MCP configuration. For live resource access, authenticate Azure CLI with the Azure account you intend to use.
Open a fresh Copilot chat and inspect available skills in the chat’s slash menu. Ask “List my Azure resource groups” and confirm that a tool-backed response refers to your intended account.
This Microsoft-documented route installs the base Azure skills globally for GitHub Copilot. It installs skills only, without the plugin hooks or bundled MCP connections, and does not install specialist plugins. Have Git available and GitHub Copilot plugin version 1.5.64-242 or later installed in IntelliJ IDEA.
In IntelliJ IDEA’s settings or preferences, open Tools → GitHub Copilot → Chat. Select Enable Skills, then Apply and OK.
Run this in your terminal. It installs the skills for GitHub Copilot across projects and accepts the installer’s prompts.
npx skills add https://github.com/microsoft/azure-skills/tree/main/.github/plugins/azure-skills/skills -a github-copilot -g -yWait for the installer’s successful-addition confirmation. In GitHub Copilot chat, use Microsoft’s documented skill entry to display the available Azure skills.
/skill:azureThese IntelliJ steps come from Microsoft’s Azure Skills installation guide. Azure accounts, tools, and permissions needed for a particular skill still apply when you use it.
Install the base azure plugin from Microsoft’s Azure Skills marketplace, then optionally add specialist plugins. For Azure MCP tools, have Node.js 22 or later with npx on PATH. For live Azure work, use an Azure identity with access to the resources involved; authenticate Azure CLI. Azure Developer CLI authentication is needed for workflows that deploy with azd.
Azure MCP downloads and runs through npx, which inherits your local .npmrc configuration. Install only on a trusted device with a trusted npm configuration: a compromised configuration could cause malicious code to run. Plugin telemetry hooks are enabled with the plugin; the opt-out is above the route chooser.
In a Copilot CLI session, add Microsoft’s marketplace once.
/plugin marketplace add microsoft/azure-skillsInstall the base Azure plugin in that session. If it is already installed, keep that copy.
/plugin install azure@azure-skillsOptionally add specialist plugins for the workflows you need. Keep exactly one copy of azure, whichever marketplace supplied it; adding a specialist does not require another base-plugin install. Run only the commands for the specialists you want.
aks-skills: AKS incident diagnosis, GPU inference, packet capture, and cluster efficiency. Run in your terminal.
copilot plugin install aks-skills@azure-skillsazure-cost: Spending analysis, estimates, optimization, and governance. Run in your terminal.
copilot plugin install azure-cost@azure-skillsazure-advisor: Advisor recommendations and remediation guidance. Run in your terminal.
copilot plugin install azure-advisor@azure-skillsazure-kusto-graph-skills: Kusto graphs, IRQL investigations, and graph visualization. Run in your terminal.
copilot plugin install azure-kusto-graph-skills@azure-skillsazure-local-skills: Standard and multi-rack Azure Local workflows. Run in your terminal.
copilot plugin install azure-local-skills@azure-skillsfoundry-iq-skills: Knowledge-base workflows. Helper scripts need Python 3.12 or later; Microsoft documents Copilot CLI support on Windows and Linux. Run in your terminal.
copilot plugin install foundry-iq-skills@azure-skillsStart a new Copilot CLI session so its hook configuration loads. In your terminal, list installed plugins and confirm that azure and each specialist you selected are present.
copilot plugin listIf you use Azure China or US Government, configure the MCP server and CLI identity for that cloud before starting the server or checking access: sovereign-cloud setup.
In the new Copilot CLI session, inspect the configured MCP servers and confirm that Azure is available. If you installed azure-cost, also check for its Azure Resource Manager MCP connection.
/mcp showAsk “List my Azure resource groups” and check that the response uses your intended Azure account. This is Microsoft’s suggested connection check.
Install Microsoft’s Azure plugin only through Cursor’s plugin interface; this route does not separately install specialist marketplace plugins. Azure MCP needs Node.js 22 or later with npx available. The package includes telemetry hooks; use the opt-out above the route chooser before launch if you do not want collection.
Azure MCP downloads and runs through npx, which inherits your local .npmrc configuration. Install only on a trusted device with a trusted npm configuration: a compromised configuration could cause malicious code to run.
Open Cursor’s Customize page or plugin marketplace panel. Search for Azure and select Microsoft’s Azure plugin.
Install it at the project or user level, according to where you want it available.
For Azure China or US Government, configure the MCP server and CLI identity for that cloud before starting the server or checking access: sovereign-cloud setup.
In Agent chat, type a slash and search for an Azure skill. This checks that the skill is available to the agent.
For live Azure access, authenticate Azure CLI, then ask “List my Azure resource groups.” Confirm that the response uses tools against your intended account.
Sources: Microsoft’s Azure installation entry for Cursor and Cursor’s plugin installation guidance.
Use Codex CLI to add Microsoft’s marketplace and install only its base azure plugin. These steps do not install specialist plugins. Plugins are not supported in the Codex IDE extension. Azure MCP needs Node.js 22 or later with npx available; resource operations need an Azure identity with the corresponding access.
Azure MCP downloads and runs through npx, which inherits your local .npmrc configuration. Install only on a trusted device with a trusted npm configuration: a compromised configuration could cause malicious code to run. Apply the telemetry opt-out above the route chooser before launch if you do not want collection.
In your terminal, add Microsoft’s Azure Skills marketplace.
codex plugin marketplace add microsoft/azure-skillsIn Codex CLI, open the plugin browser.
/pluginsOpen the azure-skills marketplace and install azure.
For Azure China or US Government, configure the MCP server and CLI identity for that cloud before starting the server or checking access: sovereign-cloud setup.
Start a new Codex session. Reopen the plugin browser to confirm that the Azure plugin is enabled, then inspect the available skills.
/skillsInspect the active MCP servers in the new Codex session.
/mcpFor live resource access, authenticate Azure CLI, then ask “List my Azure resource groups.” Confirm that the result uses your intended Azure account.
Sources: Microsoft’s Codex instructions and OpenAI’s marketplace compatibility documentation.
This installs the repository’s base Azure extension only, which exposes 28 of the catalog’s 46 listed skills and configures Azure MCP. It does not separately install the marketplace’s specialist plugins. Have Node.js 22 or later with npx available.
Azure MCP downloads and runs through npx, which inherits your local .npmrc configuration. Install only on a trusted device with a trusted npm configuration: a compromised configuration could cause malicious code to run.
Run this in your terminal, outside Gemini CLI’s interactive session.
gemini extensions install https://github.com/microsoft/azure-skillsIf you use Azure China or US Government, configure the MCP server and CLI identity for that cloud before starting the server or checking access: sovereign-cloud setup.
Restart the Gemini CLI session so the extension changes take effect.
In Gemini CLI, list extensions and confirm that Azure is installed and active.
/extensions listList the skills available to the session.
/skills listFor live access, authenticate Azure CLI, then ask “List my Azure resource groups.” Review Gemini’s skill-activation consent prompt if one appears, and confirm that the result uses your intended account.
Use APM to install the repository’s Azure package into an existing project’s detected agent environments. This installs that package only; it does not separately select the marketplace’s specialist plugins. APM accepts Claude-style plugin manifests and integrates supported resources into its selected targets. Start in a project with recognizable agent configuration; without a target signal, APM stops with a target-selection error.
Have APM installed using its official installation guide. Open a terminal at the root of your configured project. For Azure MCP, have Node.js 22 or later with npx available.
Install Microsoft’s Azure Skills package into that project.
Azure MCP downloads and runs through npx, which inherits your local .npmrc configuration. Install only on a trusted device with a trusted npm configuration: a compromised configuration could cause malicious code to run.
apm install microsoft/azure-skillsCheck the installation output for the intended targets and inspect the updated apm.yml, apm.lock.yaml, and generated target directories. Keep apm_modules available because installed resources can reference it.
For Azure China or US Government, configure the MCP server and CLI identity for that cloud before starting the server or checking access: sovereign-cloud setup.
Open a fresh session in your target AI tool and confirm that Azure skills are available through its skill browser. For Azure MCP, use an authenticated Azure CLI identity, then ask “List my Azure resource groups” and check the account used.
APM’s target rules determine where resources are integrated; do not assume every AI tool receives identical components. See APM package installation and target selection.