What is the Microsoft Foundry Control Plane?

The Microsoft Foundry Control Plane is a unified management interface that provides visibility, governance, and control for AI agents, models, and tools across your Foundry enterprise. The Foundry Control Plane serves as your central location for managing every aspect of your AI fleet, from build to production.

As organizations evolve from isolated copilots to autonomous multi-agent fleets, they need unified oversight. The Foundry Control Plane provides the visibility, governance, and control needed to scale with confidence.

The Foundry Control Plane allows you to:

• Gain unified visibility across all Foundry, Microsoft, and third-party agents in one place.
• Observe, analyze, and optimize agent performance, cost, quality, and compliance.
• Apply deterministic guardrails and policies for safety, compliance, and operational consistency.
• Visualize resource consumption, detect anomalies, and optimize agent configurations to reduce spend while maintaining reliability.

The need for unified AI fleet management

Over the next 12–24 months, enterprises will move from a handful of agents to thousands of active, semi-autonomous entities. Without unified oversight, this growth creates risk: uncontrolled costs, data leakage, task drift, and inconsistent security postures.

The Foundry Control Plane is built to address that inflection point. Without proper control mechanisms, scaling AI agents becomes risky and inefficient.

The Foundry Control Plane provides:

• Fleet-wide visibility - See all agents (Foundry-native, Microsoft, and third-party), models, and tools in one place.
• Governance at scale - Apply guardrails, enforce policies, and monitor compliance posture.
• Observability - Monitor and trace agents for debugging, remediation, and optimization.
• Optimization intelligence - Reduce cost, improve performance, and maintain reliability.
• Alerting and notifications - Receive real-time alerts when security, performance, or compliance issues require your attention, so you can act before risk escalates.
• Comprehensive guardrails - Apply a suite of built-in guardrails to models and agents to protect your AI systems against unsafe behaviors, data leakage, and prompt injection attacks.

Core Functionalities

The Foundry Control Plane consolidates inventory, observability, compliance, and security into one role-aware interface. It integrates seamlessly with Microsoft security and governance systems (Defender, Purview, Microsoft Entra) to deliver trust at scale.

Core capabilities include:

Unified Fleet Visibility

Get a real-time overview of every agent, model, and tool across projects in your subscription.

• Track key performance indicators such as active agents, run completion, compliance posture, cost efficiency, and prohibited behaviors.
• Use deep links to evaluation and monitoring experiences for rapid debugging, diagnosis, and remediation.
• Visualize fleet health through intuitive dashboards that surface trends and anomalies instantly.

Observe to Protect and Improve

Observability is built in, not added on.

• Correlate alerts, evaluation results, and trace data to pinpoint issues instantly.

• Continuously evaluate agent performance, quality, and risk dimensions such as Task Adherence, Intent Resolution, Tool Call Success, Groundedness, Sensitive Data Leakage, and Jailbreak/XPIA exposure.

• Use the AI Red Teaming Agent and cluster analysis for automated vulnerability probing and error root-cause discovery.

• Let Foundry Agent recommend improvements, from prompt refinements to model version upgrades.

Govern and Enforce with Guardrails

Operationalize Responsible AI principles through policies that scale.

• Define enterprise-wide guardrail policies for safety, compliance, and quality.

• Apply bulk remediation to instantly correct noncompliant configurations across your fleet.

Secure Agents

• Schedule automated red-teaming scans and drift monitoring for ongoing agent testing.

• View Defender and Purview alerts directly in the Control Plane dashboard.

• Track rate limits, token usage, and cost anomalies to prevent inefficiency or abuse.

Key features

The Foundry Control Plane experience begins in the Operate tab, your command center on the upper right-hand side of the Foundry workspace. From Operate, you can monitor, govern, and optimize every agent, model, and deployment within your subscription. Each sub-tab within Operate is designed around a specific job-to-be-done (JTBD), helping different roles, from builders to administrators, manage AI systems confidently at scale.

Overview

Use this page to understand fleet health, performance, and compliance at a glance.

The Fleet Overview page provides a high-level snapshot of your AI estate, aggregating key operational and compliance metrics in one view.

• View key stats such as active agents, cost trends, run completion rate, and prevented behaviors.
• Drill into anomalies or cost spikes through contextual charts and direct links to Inventory, Observability, or Policy pages.
• Identify potential risks early with trend-based health scores and alert summaries.

Assets

Use this view to track, analyze, and manage every agent, model, and tool from one place.

The Inventory view provides a unified, searchable table of all AI assets across projects within a subscription. It brings together critical metadata and health indicators, so you can assess and act on your AI estate efficiently.

• Filter and sort by key attributes such as version, tags, health score (%), cost, alerts, and token usage to locate assets quickly.
• Drill down from any entry in the Agent Inventory table into Evaluation or Monitoring tabs for pre- and post-deployment insights.
• Surface inline recommendations to refine prompts, upgrade models, or optimize configurations based on performance and cost signals.
• Correlate runtime logs with evaluation results to uncover root causes of errors or performance degradation.
• Visualize drift, latency, and error clusters across runs or builds to detect emerging issues early.
• Integrate with the AI Red Teaming Agent to automate vulnerability probing, regression testing, and issue reproduction.
• Observe and modify model and agent guardrails.

Together, these capabilities turn the Inventory into the operational backbone of the Control Plane, a single pane to understand, improve, and secure every AI asset in your environment.

Compliance

Use this tab to govern your AI systems and enforce the right guardrails.

The Policy & Security tab empowers organizations to define, apply, and continuously monitor guardrails and compliance policies across their AI estate. It provides a unified interface to operationalize Responsible AI principles while ensuring enterprise-grade safety and regulatory alignment.

• Define and enforce protections through deep integrations with Azure Policy, Microsoft Defender, and Purview, ensuring that identity, data, and threat safeguards work in concert.
• Apply versioned policies and track assignments to maintain full auditability and traceability across agents and environments.
• Monitor compliance posture in real time, surfacing noncompliant assets and enabling bulk remediation directly from the Control Plane.

Policy Management in Foundry allows administrators and developers alike to embed quality and safety requirements into the development and deployment lifecycle. These policies ensure that all models operate safely, adhering to organizational and regulatory standards.

Quota

Use this tab to view, adjust, and request quotas.

The Quota tab allows customers to easily see their model deployments and how much quota each deployment is consuming. It gives insights into usage patterns and helps manage resources effectively.

Admin

Use this tab to view, organize, and administer all projects, users, and connected resources across your Foundry environment.

The Admin tab extends your operational view beyond a single project. While most work in Foundry takes place within a project context, Manage provides an enterprise-level lens to oversee and configure multiple projects, user permissions, and linked Azure resources from one place.

From Operate → Manage, administrators and power users can:

• Gain visibility into all projects across their subscription or tenant, including active, inactive, and archived workspaces.
• View detailed project information such as owners, region, connected services, and compliance posture.
• Add or remove users directly from a project, defining granular access levels aligned with organizational roles.
• Attach or manage connected resources, such as storage accounts, compute clusters, and AI services, ensuring projects remain properly provisioned.
• Assign access at parent scope (subscription or resource group) to apply consistent governance and permissions inheritance across multiple projects.

Together, these capabilities make Manage the Control Plane's administrative backbone. It's a centralized console to ensure every Foundry project remains properly configured, compliant, and connected to the right people and infrastructure.

Related content

The Foundry Control Plane brings together everything you need to manage, govern, and optimize your AI estate, from identity and observability to policy and performance. Now that you understand the core concepts and key capabilities, explore the following guides to go deeper into setup, usage, and optimization.

Quickstarts - Get up and running quickly with these introductory guides:

• Create a Guardrail Policy - Define your first guardrail policy and understand how guardrails apply across projects.

How-to guides - Deepen your operational expertise with hands-on workflows:

• Monitor Fleet Health and Performance - Track key health metrics, alert trends, and anomalies across agents, models, and tools.
• Ensure Compliance and Security - Learn how to enforce Responsible AI policies, integrate Defender and Purview signals, and respond to compliance alerts.
• Optimize model cost and performance - Analyze cost drivers, token usage, and resource consumption to achieve higher ROI from your agent fleet.
• Register Custom Agents - Extend your governance surface by bringing third-party or external agents into the Foundry Control Plane registry.