On May 1, 2026, Microsoft made generally available two related products that had been building through the spring: Agent 365, a unified control plane for enterprise AI agent governance, and Microsoft 365 E7 — branded the “Frontier Suite” — the first enterprise license tier to bundle productivity, identity, Copilot, and agent governance into a single SKU.
The timing is deliberate. Across enterprises, AI agents are proliferating faster than IT can track them. Copilot Studio agents, third-party agents, locally running models — agents that act on behalf of users, have access to company data, and can invoke tools are becoming an infrastructure problem, not just a feature question. Agent 365 is Microsoft’s answer to that problem.
What Agent 365 Is
Agent 365 is a control plane — not an agent itself. It does not write code or run tasks. What it does is give administrators a single place to see, control, and secure every AI agent operating in their tenant.
The product is built on three pillars:
1. Observe
Agent 365 maintains a centralized agent registry — a unified inventory of every agent active in your environment, including agents built with Microsoft tools (Copilot Studio, Azure AI Foundry) and agents from ecosystem partners through registry sync. For each agent, admins can see:
- What identity the agent runs under
- Which users have access to it
- Which tools and MCP servers it is permitted to call
- Adoption and activity metrics
- Health and anomaly signals
This is the part that matters most in the first six months of deployment. Most enterprises discovering Agent 365 will find they have more agents than they thought — shadow AI at the agent layer is real, and visibility is step one.
2. Govern
Governance in Agent 365 works through policy templates applied at onboarding. When a new agent is registered (either by an admin or synced from an external platform), security policies attach automatically: access controls, data exposure limits, lifecycle rules.
Key governance features:
- Rules-based lifecycle management: inactive agents are flagged or expired automatically; ownerless agents are flagged for review
- Least-privilege access control: admins define which data, tools, and other agents each agent can reach — enforced rather than advisory
- Shadow AI detection: Intune integration lets admins detect and block unsanctioned agents running locally on managed endpoints
The lifecycle automation is particularly practical. Without it, agent proliferation leads to orphaned agents with stale access — a compliance and security liability.
3. Secure
Security in Agent 365 extends Microsoft’s existing enterprise defense stack to agent identities:
- Microsoft Entra enforces conditional access and risk-based authentication for agents acting on user behalf
- Microsoft Purview applies DLP and information protection policies to data agents can access or produce
- Microsoft Defender provides continuous behavioral monitoring, anomaly detection, and real-time threat response for agent activity
The framing Microsoft uses here — treating agents as a new class of identity, not just a new class of application — is the right model. An agent with access to your CRM and email is a privileged identity. Entra conditional access is the appropriate control surface.
Microsoft 365 E7 — The Frontier Suite
Agent 365 is available standalone at $15 per user per month, but its launch is tied to the broader E7 bundle, which packages:
| Component | Standalone price |
|---|---|
| Microsoft 365 E5 | $60/user/month |
| Microsoft 365 Copilot | $30/user/month |
| Microsoft Entra Suite | $12/user/month |
| Agent 365 | $15/user/month |
| Total à la carte | $117/user/month |
| E7 Frontier Suite | $99/user/month |
The bundle saves $18/user/month — about 15%. At 1,000 users, that is $216,000 per year. At 5,000 users, it is over a million dollars annually.
The question of whether E7 is worth it depends entirely on which components you actually need. If you are running E5 without Copilot and have no near-term plans to deploy AI agents at scale, the jump from E5 ($60) to E7 ($99) is a 65% price increase for capabilities you are not yet using. If you are already on E5 plus Copilot and are beginning to manage agent sprawl, the math tips clearly in E7’s favor.
The E5 price also drops to $60/user from July 2026 (down from ~$66 previously), which narrows the E5-only baseline and makes E7 appear slightly more expensive in relative terms — but the absolute savings on the full stack still hold.
Who This Is For
Large enterprises with active Copilot deployments. If you have rolled out Copilot across thousands of seats, agents are already proliferating in your tenant. Agent 365 gives security and compliance teams the visibility and control they need to manage that safely.
Regulated industries (financial services, healthcare, government). The Purview and Defender integration makes Agent 365 the clearest path to agent governance that satisfies audit requirements. The agent registry and activity logging are exactly what compliance teams will ask for when an AI agent touches sensitive data.
Organizations planning multi-agent architectures. If your roadmap includes custom agents built on Azure AI Foundry or Copilot Studio, getting the governance infrastructure in place before deployment is far easier than retrofitting it.
IT and security teams building AI policy. Agent 365’s policy templates and lifecycle automation give IT a structured way to enforce whatever governance rules the organization decides on — without relying on agent developers to self-police.
What It Does Not Do
Agent 365 is not a development platform. It does not help you build agents. That’s Copilot Studio and Azure AI Foundry.
It is also not a runtime. Agents do not run inside Agent 365 — they run on their own infrastructure and are registered with, monitored by, and governed through Agent 365.
Third-party agent coverage depends on registry sync support from the external platform. At GA, a set of ecosystem partners are supported; coverage will expand, but if you are running agents from a vendor who has not integrated with Agent 365 yet, you will have visibility gaps.
The local agent shadow AI detection is early-stage. Detection relies on Intune coverage of managed endpoints; unmanaged devices are out of scope.
Verdict
Agent 365 solves a real and growing problem: enterprises do not have a coherent way to manage AI agents as a population of identities with access to sensitive systems. The observe-govern-secure framework is sound, the Entra and Defender integration is mature, and the agent registry concept — borrowed from how mature organizations manage service accounts — is exactly the right mental model.
The $15/user/month standalone price is reasonable for large organizations with active agent deployments. The E7 bundle makes clear financial sense if you are already running or seriously planning Copilot plus third-party agents. If you are still evaluating whether to deploy AI at scale, the bundle is premature.
Rating: 4/5. Addresses a real enterprise need with enterprise-grade tooling. Docked one point for early-stage local agent detection and third-party coverage gaps that will take 12-18 months to close.
ChatForest covers AI tools and infrastructure. This review is based on published Microsoft documentation and third-party analysis; we have not independently tested Agent 365 in a production environment.