What problem does it solve? Automating Blender work requires choosing between a live GUI session and headless processes, handling undo steps, context overrides, and verifying that changes actually happened. This Skill routes every job to the right tool across two MCP servers and a set of scripts, so scene edits, renders, conversions, and drawings run deterministically without disturbing the user's open Blender window. ## Core Features & Use Cases - Live and headless execution: Run code in the user's Blender GUI through execute_blender_code with automatic undo steps, or in factory headless sessions via headless.py for long jobs, batches, and renders. - Verification with snapshots and captures: Record evaluated scene state as JSON snapshots and framed viewport or Workbench renders as PNGs, with pixel diffs against earlier captures to prove each change. - Domain workflows: Reference guides cover rendering, look development, animation, drafting sheets at scale, geometry nodes, CAD interchange, IFC/BIM authoring with Bonsai, and geospatial imports. - Use Case: Convert a batch of STEP and IFC files to glTF in a headless session, then snapshot and capture the imported geometry to confirm unit scale and object counts before delivering the files. ## Quick Start Ask the agent to open your current Blender scene, add a 2 meter cube at the origin, and show a captured front view with a snapshot of what changed.
