A walkthrough of building a Blender-powered AI agent hosted in Microsoft Azure AI Foundry and exposed through Microsoft 365 Copilot. The demo shows running Blender in headless mode inside a Docker container, wrapping it as a Python agent using tools adapted from Blender MCP, and deploying it to the Foundry hosted agent service. The workflow covers local debugging in VS Code with the AI Toolkit extension, deploying to Foundry with configurable CPU/memory resources, storing rendered outputs in Azure Blob Storage, and monitoring agent behavior via Azure Monitor telemetry including tool call frequency and token consumption. The end result lets users generate and iterate on 3D scenes through natural language in M365 Copilot, then download the resulting Blender files for further editing.
Sort: