Microsoft adds cost management and pricing tools to Azure Resource Manager MCP server

Microsoft has extended its Azure Resource Manager MCP server with cost management and pricing capabilities, allowing AI agents to query cost data, retrieve pricing, and manage budgets within cloud workflows. The update, released by Microsoft Cost Management, adds a set of tools that let language model agents combine Azure resource information with cost and optimization insights.

The Azure Resource Manager MCP server is a Model Context Protocol (MCP) implementation that gives AI agents programmatic access to Azure's control plane. With the addition of two new toolsets—Cost Management and Pricing—agents can now perform tasks such as querying historical spend, forecasting future costs, creating budgets and alerts, retrieving retail or negotiated prices, and analyzing Azure Advisor savings recommendations. The tools also support AKS-specific cost queries by cluster and namespace.

According to Microsoft, the goal is to reduce the gap between cloud operations and cost data, which is often scattered across separate tools. Rather than switching between the Azure portal, cost analysis dashboards, and pricing calculators, a developer or operator could ask an AI assistant to estimate the cost impact of changing a VM size in an ARM template, or to show the top cost drivers for the month. Microsoft provides example prompts such as "What were my top cost drivers this month?" and "Estimate the monthly cost of a three-node AKS cluster with my negotiated pricing."

The examples illustrate potential workflows, but the actual capabilities are defined by a set of tools exposed via MCP. The Cost Management toolset includes functions for querying costs (query_costs, query_aks_costs), forecasting, listing dimensions, managing budgets and alerts, and retrieving reservation and savings plan recommendations and utilization. The Pricing toolset adds functions to get retail prices and download negotiated price sheets.

Access to the new tools is opt-in. Users must have Visual Studio Code installed, a valid Azure account with appropriate permissions, and a GitHub Copilot subscription. After installing the Azure Resource Manager MCP server, the cost tools are enabled by adding specific x-mcp-toolset headers to the client configuration—either at the workspace or user level in VS Code, or via the Copilot CLI configuration file. The server must be restarted afterward.

The release comes with limitations typical of such integrations. The cost data agents can access is only as accurate and timely as Azure's own cost management backend. Budgets and alerts are managed through the same Azure APIs, so agents are bound by existing data granularity and update latencies. Furthermore, the tools require both MCP server and client support, which currently centers on VS Code and GitHub Copilot.

Microsoft notes that additional tools and skills are planned, including best-practice guides for advanced cost scenarios, but those are not yet available. For now, the release gives developers a way to experiment with cost-aware agent interactions, provided they configure the headers and understand that agents query data rather than performing autonomous actions beyond what the toolset exposes.