Using Meta XR Operator with Meta XR Simulator
Updated: Jul 23, 2026
Experimental feature
Meta XR Operator is an experimental component of the Meta XR Core SDK. Tool names, APIs, and behavior may change in future releases. Avoid depending on it in production apps. Meta XR Simulator runs your XR app on your computer without a headset. Combined with Meta XR Operator, it’s the fastest way to let an AI agent drive and inspect your app. Because Meta XR Operator works at the OpenXR layer, the same MCP tools work whether you run in the Simulator, in the Editor with Quest Link, or on a headset.
- Iterate quickly without headset hardware or a tethered device.
- Simulate mixed reality, including passthrough and spatial entities, so tools that read the room work.
- Run repeatable, agent-driven development and testing sessions on a single machine.
Enter Play mode. The Simulator window opens and your app runs in XR. Meta XR Operator starts its local MCP server inside the running app.
Connect your agent and ask it about the running app. Your agent connects through the registered MCP proxy, so you don’t work with the underlying endpoint directly. See
Getting started for more information.
Simulate a mixed reality environment
To let an agent reason about the physical space, such as listing rooms, walls, and furniture, run your app against a synthetic environment that provides mixed reality data. With a synthetic environment loaded, an agent can query spatial entities and capture composited images through Meta XR Operator, just as it would on a headset. See the
Meta XR Simulator overview for how to enable and configure synthetic environments.
Control the simulated session
Meta XR Operator lets the agent set the head and controller poses and send controller input, which is especially useful in the Simulator where there’s no physical headset to move. You can also use the Simulator’s own keyboard and mouse controls to move the view and trigger input alongside the agent.
- Some tools behave differently in the Simulator than on a headset.
- Image capture returns simulated imagery, not real passthrough video.
- Spatial entities reflect the synthetic environment you load, not a real room scan.