News
Microsoft Formalizes AI Agent Stack with Identity, Collaboration, and Orchestration Tools at Build 2025
Microsoft introduced a set of integrated tools and frameworks at its Build 2025 conference designed to make AI agents more secure, manageable, and operational across enterprise environments. The updates span identity governance, multi-agent coordination, model deployment, and developer automation, reinforcing Microsoft’s strategic shift toward treating agents as core digital entities in the software ecosystem.
The announcements included the public preview of Microsoft Entra Agent ID, support for multi-agent systems in Copilot Studio, new production capabilities in Azure AI Foundry, and a GitHub Copilot agent for automated development tasks.
Agent Identity Management Added via Microsoft Entra Agent ID
A centerpiece of the launch was the Microsoft Entra Agent ID, a new identity and governance layer for nonhuman software agents. The service allows IT teams to apply visibility, access control, and lifecycle policies to agents built using Microsoft services such as Copilot Studio, Azure AI Foundry, and Security Copilot.
Through the Entra admin portal, enterprises can apply Conditional Access, auditing, and lifecycle rules to agent identities. Microsoft also announced plans to support third-party agent integration and is collaborating with ServiceNow and Workday to extend governance into multi-vendor workflows.
Copilot Studio Adds Support for Multi-Agent Systems
Copilot Studio, Microsoft’s low-code agent development platform, now supports multi-agent orchestration, allowing developers to build agents that delegate tasks to other agents. This enables coordinated workflows across services such as Microsoft 365, Azure AI, and Microsoft Fabric. Current use cases in private preview include CRM data processing, document drafting, and meeting scheduling.
Other Copilot Studio updates include:
- Software-controlling agents capable of interacting with user interfaces.
- Custom model integration via Azure AI Foundry.
- Expanded publishing endpoints, including SharePoint and WhatsApp (expected July).
- Policy enforcement tools, including DLP and Entra integration.
- Agent routing suggestions, matching user queries to appropriate agents.
Azure AI Foundry Launches with Agent-Centric Enhancements
Azure AI Foundry, now generally available, adds a robust agent operations layer with support for the Model Context Protocol (MCP) and Agent-to-Agent (A2A) communication. The platform enables scalable agent deployment with native orchestration and monitoring tools.
Key features include:
- A catalog of 1,900+ models, including GPT-4.5 and Grok 3.
- Support for edge deployments via Foundry Local.
- A new developer tier with free hosting.
- Dynamic model routing for cost-performance optimization.
- Built-in AgentOps tools for debugging and monitoring.
Foundry integrates with Visual Studio Code and supports both individual agents and multi-agent topologies through a unified API.
GitHub Copilot Adds Autonomous Coding Agent
GitHub also introduced a new Copilot coding agent capable of autonomously resolving GitHub issues by generating code and submitting pull requests. The agent operates within GitHub Actions with safeguards for code review and branch enforcement.
Features include:
- Real-time session monitoring.
- Visual input support and MCP compliance.
- Repository-level security policies.
The Copilot agent complements GitHub’s broader portfolio of AI-driven coding tools, expanding from completion and chat into automated task execution.
Microsoft’s announcements mark a transition toward agent-first infrastructure across its platforms. From secure identity provisioning and multi-agent orchestration to production-ready hosting and development automation, the company is positioning AI agents as composable, governed, and integrated components of the enterprise stack.
Let me know if you'd like a short version for enterprise IT briefings or a diagram showing how Microsoft’s agent ecosystem fits together.