mcp-for-blender-pro: talk to your Blender scene, build geometry, swap materials and pull assets — then hand the work to an agent team

What it does

mcp-for-blender-pro is an AI assistant add-on for Blender that also ships a built-in MCP (Model Context Protocol) server. Put simply, it lets you talk to your own 3D scene in natural language: ask the AI to inspect what is in the scene, build a product-display layout, arrange objects, swap materials, or prep an export — and you watch the viewport the whole time without leaving Blender. This is not one of those tools that just spits out a render and calls it done. It actually drives Blender objects, lights, cameras, and materials to carry out the work, turning a creative brief into a chain of real scene operations.

Why it is trending

Blender is the most popular free, open-source 3D app around, yet writing scripts and wrestling with the bpy API has always had a learning curve. At the same time, AI assistants are moving toward actually operating software instead of just chatting. This add-on bridges the two: it pairs a visible chat workspace with 21 scene tools and a built-in MCP server that can hook into external clients like OpenCode. In other words, it folds conversation, scene tools, model choice, and agent workflows into one surface — exactly the pain point for artists and technical artists who want speed without leaving their viewport. On top of that, Blender's 4.2 extension system lowered install friction to almost nothing, so a tool like this can reach ordinary artists, not just coders.

Technical highlights

The parts that genuinely earn their keep:

• 21 scene tools. They cover creation (add_cube, add_sphere, add_light, add_camera, and more), editing (move, duplicate, delete, set_material), visual feedback (viewport_screenshot, execute_code), API lookup (bpy_api_lookup, describe_node_type), and export (export_scene). Basically every daily modeling action is wrapped.
• Multi-agent teams. You can configure a small crew — a Manager breaks the request into tasks with acceptance criteria, Workers perform the scene operations, and Reviewers inspect results and either approve or return specific findings. Tasks run in order so scene edits stay coherent; a blocked task gets one correction attempt, and a second failed review stops the workflow.
• A visual Studio window. Chat, scene objects, viewport snapshots, approvals, and a task board all live in one companion window, so you are not jumping between panels. The UI itself is multilingual: the main menu switches between English, German, French, Spanish, and Chinese — and the companion Studio is where the agent team really shines: you assign each role a responsibility, instructions, and a model, then watch assignments, attempts, and review findings populate a task board as work progresses.
• Your choice of model. It supports OpenAI, Claude, Gemini, OpenRouter, Ollama, and custom endpoints — cloud or local — and lets you save multiple connections and keep work and personal keys in separate slots, including multiple keys for the same provider.
• Asset integrations. Poly Haven, Sketchfab, and Poly Pizza are wired in, so it can search resources and ask before importing.

A nice touch is the change governance: the built-in chat supports operation approvals, cancellation after the current step, and named undo steps for mutations. Python tool validation catches disallowed operations — though it is not a security sandbox, and external MCP clients manage their own approval policies. The docs also hand you great starter prompts, like inspect this scene and explain its object structure before making any changes, or create a simple product-display scene with a pedestal, a camera and three lights, and give the pedestal a matte blue material.

Who it is for

It fits 3D artists, technical artists, developers who want to connect AI into Blender, and anyone who enjoys tinkering with AI workflows. If you do repetitive modeling, layout, or material work, or you want to rough out a scene from a sentence and then refine it by hand, it saves real time. For developers, the built-in MCP server means you can plug Blender into a larger agent pipeline through a client like OpenCode.

Quick try

Download mcp-for-blender-pro.zip from the project Releases page. In Blender open Edit → Preferences → Get Extensions → menu → Install from Disk, pick the ZIP, enable mcp-for-blender-pro, then switch to the 3D Viewport, press N, open the MCP Pro panel, choose a model, and start chatting. It targets Windows x64 with Blender 4.2 or later. A good first prompt, straight from the docs: inspect this scene and explain its object structure before making any changes.

How it compares

• Versus chat-only AI (a web ChatGPT handing you a bpy script): this add-on has the assistant act directly inside your scene, so you see results live and can tell it to adjust, instead of copying a script and running it yourself.
• Versus older Blender automation scripts: the multi-agent crew chains understand intent, execute, review together, which makes complex tasks steadier and more explainable.
• Versus sourcing assets by hand: the built-in Poly Haven, Sketchfab, Poly Pizza integration lets the agent find and import resources with your permission, instead of you hunting across sites.
• A small note: several Blender MCP add-ons have appeared recently with similar ideas. When picking one, look at tool coverage, local-model support, and whether there is an approval or undo mechanism — this one covers all three fairly completely.