Microsoft Agent 365 Hits General Availability as Enterprise AI Agent Control Plane
Microsoft's control plane for AI agents went GA on May 1 at $15 per user per month, adding multicloud discovery, shadow-AI detection on Windows endpoints, and runtime blocking of malicious agent behavior.
Microsoft on May 1, 2026 declared Agent 365 generally available, ending the platform's preview period and giving enterprises a single control plane to "observe, govern, and secure" the rapidly proliferating population of AI agents inside their organizations. The standalone offering ships at $15 per user per month and is bundled into the new Microsoft 365 E7 license — the company's "Frontier Worker Suite" — which packages Agent 365 with Microsoft 365 Copilot and Microsoft's full security stack at $99 per user per month.
Agent 365 treats every agent — whether it lives in Copilot Studio, on a Windows endpoint, or inside another vendor's cloud — like an identity Microsoft already governs. The GA release adds Microsoft Defender and Intune-powered discovery of "shadow AI" agents on local devices, an asset map that ties each agent to its associated identity and cloud resources, and policy-based runtime controls that can block malicious coding agents mid-execution before they reach production. Network controls now extend across Copilot Studio agents and endpoint-based agents alike, and admins can throttle risky web connections or unsanctioned model usage from a single console.
The launch leans hard into multicloud reach. Connectors for Amazon Bedrock and Google Cloud, both in public preview, let admins automatically inventory agents that were never spun up inside Microsoft's stack. Ecosystem integrations went live with Genspark, Zensai, Egnyte, Zendesk, Kasisto, Kore.ai, and n8n, while local-agent support expanded to OpenClaw with GitHub Copilot CLI and Claude Code on the roadmap. A separate public preview, Windows 365 for Agents, runs agents inside policy-controlled Cloud PCs managed through Intune, giving them the same identity, security, and lifecycle controls already used for human employees.
"Identity, security, and governance are built in from the start," said Kore.ai CEO Raj Koneru, framing the GA as the moment enterprises can finally stop piloting agents and start scaling them. NTT DATA's Yuji Shono echoed the message, saying organizations can now "scale and govern AI agents with confidence" without sacrificing enterprise-grade security.
The pitch lands as boards grapple with what Microsoft has elsewhere called agent sprawl — thousands of unmonitored bots making API calls, signing contracts, and shipping code across hybrid environments. Agent 365 reframes that sprawl as a manageable identity problem and bets that whoever owns the control plane for agents owns the next decade of enterprise IT spend. The $15-per-seat price tag undercuts most third-party agent governance tools, and the bundling into E7 telegraphs Microsoft's intent to make agent management a default line item in the next round of enterprise license negotiations.
