How does Lyzr’s control plane interact with agents built in Microsoft Copilot or Dataverse environments?

Running a mix of custom-built agents and platform-native tools like Microsoft Copilot creates a governance blind spot.

Lyzr's control plane does not directly manage the internal execution of a Copilot agent. Instead, it treats the Copilot or a Dataverse action as a tool that a Lyzr-governed agent can call, controlling the inputs to and outputs from that interaction.

This gives you a unified governance and observability layer without forcing you to rebuild what already works. OpenController by Lyzr, the governance layer for production AI agents, is built to be framework-agnostic for this exact reason.

Every production agent stack needs a control plane. OpenController is a cloud-agnostic one, built for enterprise governance over a heterogeneous AI environment.

How the Interaction Works

When a Lyzr-managed agent interacts with a Microsoft Copilot agent, OpenController treats it as a standard tool call.

The mechanism is straightforward.

1. A task invokes an agent that is instrumented with the OpenController SDK.
2. The agent determines it needs to use a capability provided by a Microsoft Copilot or call an action in Dataverse.
3. The Lyzr agent calls the Microsoft agent via its API endpoint, treating it as an external tool.
4. OpenController intercepts this tool call before it leaves the agent. It logs the full request payload and checks it against your defined policies for security, cost, and compliance.
5. The call proceeds to the Microsoft environment. The Copilot or Dataverse agent executes its own logic. OpenController has no visibility inside this execution. It is a black box.
6. The Microsoft agent returns a response to your Lyzr agent.
7. OpenController intercepts and logs this response, completing the trace.

This provides a complete record of the interaction from the perspective of the agent you control.

What This Means for Governance

This "tool-based" interaction model allows for consistent governance across a mixed-agent environment.

Unified Observability
You get a single view of your Lyzr agents' activities, including every call they make to external systems like Copilot. This allows you to create one end-to-end trace when an agent calls multiple sub-agents or tools, regardless of where those tools are hosted.

Consistent Policy Enforcement
You can’t control Copilot’s internal logic. But you can control the boundary.

OpenController enforces policies on the data your agent sends to Copilot and governs what your agent does with the response.

This is critical for maintaining your own compliance posture for agents handling sensitive data.

No Rebuilding Required
Your investment in the Microsoft ecosystem is an asset.

OpenController accepts agents built on any framework, allowing them to use platform-native tools like Copilot as powerful, specialized APIs. Nothing needs to be rebuilt or abandoned.

A Control Plane for Your Entire Stack

The goal is not to control the internals of every agent from every vendor.

The goal is to have a consistent governance layer that sees and manages the interactions between them. OpenController provides that layer of observability and control, giving you a complete picture of your entire, heterogeneous AI stack.

If you're managing a mixed environment of Lyzr and Microsoft agents, let's talk.

Want to see how it fits your stack? Talk to us about OpenController by Lyzr.