
Key Takeaways
- Zenity brings comprehensive security coverage across the Microsoft ecosystem, so security teams get one place to secure AI agents in code, in Office, and in custom builds.
- Security teams can see what agents are doing and trace actions back to a real person, even when GitHub settings hide actor names.
- Zenity stops risky actions where the platform allows, with inline prevention for Claude Enterprise in Microsoft 365, the GitHub Copilot coding agent, and, in a more limited form, Copilot Studio agents.
- High-fidelity alerts flow into Microsoft Sentinel, so the SOC triages agent issues alongside the rest of its incident queue.
Microsoft has put AI agents into nearly every part of its ecosystem, and enterprises are using them, which makes AI agent security a problem that spans every team. Developers use GitHub Copilot, business teams build in Copilot Studio and Agent Builder, employees hand work to Cowork and to Claude Enterprise in Office, and Engineering teams build custom agents in Foundry. Each one can act on company data, and each one keeps its own records.
Zenity brings comprehensive security coverage across the Microsoft ecosystem. Security teams get one place to secure AI agents in code, in Office, and in custom builds, instead of a separate view for every tool.
That coverage gives security teams answers: what agents are doing, who is behind them, what to stop, and what needs an analyst’s attention. Here’s how it works for each kind of agent. For what’s new, see Zenity’s latest announcement.
Coverage Across the Microsoft Ecosystem
Each Microsoft product gives security tools a different level of access. Zenity provides telemetry and findings across all of them and stops risky actions where the platform allows.
Coding agents in GitHub Copilot
For the coding agent in GitHub Copilot in GitHub Enterprise, Zenity provides telemetry, findings, and inline prevention. Where enterprise settings hide actor names, Zenity uses Microsoft Entra ID to resolve them, so the security team can see whose agent made a change.
Agents built by citizen developers and IT
For agents on the Microsoft Copilot harness in Microsoft Copilot Studio, Zenity provides telemetry and limited inline prevention. Zenity also extends telemetry and findings to the agents that employees create with Agent Builder in Microsoft 365 Copilot, so user-built agents stay visible.
Agents that work for employees
Newly available Microsoft Copilot Cowork transcripts let Zenity produce telemetry and findings on what Cowork does for users, so agents that act on an employee’s behalf leave a record security can use. For Claude Enterprise in Microsoft 365 apps, Zenity provides telemetry, findings, and inline prevention, so security teams can approve its use across Office with controls in place.
Custom agents in Microsoft Foundry
AI infrastructure teams build custom agents in Microsoft Foundry. Zenity’s integration has fully migrated to the new Microsoft Foundry Agent Service, so those teams stay covered on Microsoft’s current agent platform.
Agent management in Microsoft Agent 365
For organizations that manage their agents in Microsoft Agent 365, Zenity’s runtime risk signals appear directly in Agent 365. The integration covers agents built on Microsoft Copilot Studio, Microsoft 365 Copilot, and Microsoft Foundry. Zenity adds behavioral context: whether a tool call makes sense for that agent, whether a retrieval pattern matches sanctioned use, and whether a multi-step interaction holds together. When a team needs more detail, it can go from Agent 365 into Zenity for the full investigation, posture management findings, and a complete view of how each agent is configured and what it can reach.
From Coverage to Answers
Coverage matters because of what it lets security teams do with it.
See what agents are doing. Telemetry and findings give security teams one view of agent activity in code, in Office, and in custom builds.
Trace actions back to a real person. Zenity ties agent actions to identities, using Microsoft Entra ID where GitHub settings hide actor names.
Stop risky actions where the platform allows. Inline prevention covers Claude Enterprise in Microsoft 365 and the GitHub Copilot coding agent, and applies in a more limited form to agents on the Microsoft Copilot harness.
Get the alerts that matter into Microsoft Sentinel. Zenity’s upgraded Microsoft Sentinel integration replaces raw-log export with high-fidelity alerts, so security operations center (SOC) analysts triage agent issues alongside the rest of their incident queue.
Take a coding agent that changes a repository at a company whose GitHub settings hide who started it. Zenity resolves the actor through Microsoft Entra ID, records what the agent did, and produces a finding when the activity warrants attention. The analyst sees a named person, a clear account of the action, and a high-fidelity alert in the Microsoft Sentinel queue they already work.
Secure AI Agents Across Microsoft With Zenity
As your company adopts AI agents across Microsoft, Zenity helps you secure them. Visiting Microsoft Ignite? Stop by Zenity at booth #6352L to see it in action.
Can’t make it? Schedule a demo to see how Zenity brings AI agent security to the agents your teams already use.
All ArticlesRelated blog posts

Securing the Next Control Plane for AI Agents: Introducing Zenity for Sandboxes
Today we're introducing Zenity for Sandboxes: agent-aware security for every sandbox your agents run in. AI agents...

Zenity Brings AI Agent Security Leadership to the ServiceNow Control Tower Alliance
Zenity pioneered AI Agent Security because we recognized early that agents would require a different approach to...

Meet Dai Fujimoto: Building Zenity’s Next Chapter in Japan
Japan’s Ministry of Finance reports that AI adoption in large local enterprises has surged to 89%. Companies are...
Secure Your Agents
We’d love to chat with you about how your team can secure and govern AI Agents everywhere.
Get a Demo