XR AI v0.1.0 Beta Release

XR AI — v0.1.0 Public Beta

First public release. This is a beta — APIs, configs, and behavior will change.
Report issues at https://github.com/NVIDIA/xr-ai/issues.

---

XR AI is a developer stack for building multimodal, real-time AI agents for XR.
Agents see and hear the user's environment, reason over live spatial context, call
tools via MCP, and respond with audio or data — across web, iOS/visionOS, Android,
AR glasses, and XR headset clients.

What's included:

• XR Media Hub — real-time media routing with per-participant isolation
• Agent SDK (xr-ai-agent, xr-ai-models, xr-ai-pipecat)
• GPU-backed AI services: VLM, LLM (×3), STT, TTS
• Six MCP servers: scene rendering, OpenXR spatial, vector math, vision, video, transcript
• Two end-to-end samples: simple-vlm-example and xr-render-demo
• Client samples for web, web-XR, Android, iOS/visionOS, and native C++
• CloudXR integration layer

Full documentation: https://nvidia.github.io/xr-ai

---

Known limitations

• Android XR passthrough camera and immersive session APIs not yet wired
• CloudXR requires a separate license; not included
• Breaking changes expected throughout beta