blender-agent-control

Blender automation for modeling and rigging uses the blender-agent MCP to provide deterministic, context-driven control of a live session. It applies a see-and-refine loop with vision feedback and safety checks to execute precise, repeatable steps across your workflow.

The see-and-refine loop in Blender automation is a deterministic operating cycle where you call vision_snapshot after every mutation to verify results visually. This feedback mechanism ensures your modeling, lighting, or rendering adjustments are accurate before proceeding to the next step.

To execute deterministic rendering and export workflows in Blender, you use the categorized tool map for render and export commands provided by the blender-agent. After issuing commands, you call vision_snapshot to verify results and apply safety checks before finalizing the output.

Yes, Blender automation supports context-driven material and shader creation through the blender-agent's categorized tool map. You issue deterministic tool calls for materials and shaders, then verify the results using vision snapshots within the see-and-refine loop.

Limitations of using blender-agent for 3D animation control include the strict requirement to run blender_launch at session start and follow the see-and-refine loop. The skill also recommends loading aaa-modeling for complex tasks, indicating potential limits in handling highly intricate operations natively.