Blender gains multi-agent tools

A Blender assistant project is bringing chat and AI-driven scene work into the same creative workspace. The GitHub page describes tools for inspecting scenes, creating and editing objects, changing materials, finding assets and exporting scenes. Users can follow the assistant’s actions in a conversation and check results in a viewport preview.

The project supports cloud models from OpenAI, Anthropic and Google, along with local Ollama models and compatible custom endpoints. For larger jobs, its multi-agent workflow assigns tasks to a manager, workers and reviewers. Tasks run in sequence, and reviewers can approve the work or return findings for correction. The repository also describes approvals, cancellation and undo steps, though it warns that its Python validation isn’t a security sandbox.

For a small studio or a team that uses Blender, the practical point is that AI can be connected to specific scene operations, not just asked for general advice. That could make repetitive edits easier to delegate, while leaving the work visible in Blender for review. It’s one example of a workflow that could be relevant to an AI command centre. The project’s GitHub page has setup details and the full list of supported connections.