Blender MCP — AI Integration Server for 3D Creation

Introduction

Blender MCP is an open-source server that bridges AI assistants and the Blender 3D creation suite through the Model Context Protocol. It allows AI models to inspect, create, and modify 3D scenes, materials, and animations in Blender using natural language commands, turning conversational AI into a 3D content creation tool.

What Blender MCP Does

• Exposes Blender operations as MCP tools that AI assistants can invoke
• Enables creating, transforming, and deleting 3D objects through natural language
• Supports material assignment, lighting setup, and camera positioning via AI commands
• Provides scene inspection tools so AI can understand the current 3D workspace
• Allows executing arbitrary Blender Python scripts for advanced operations

Architecture Overview

Blender MCP runs as a server inside Blender via a Python addon that listens for MCP protocol messages. The addon registers a set of tools covering object manipulation, scene queries, material management, and script execution. When an AI client sends a tool call, the addon translates it into Blender Python API calls executed in the main thread. Results and scene state are serialized back to the client for context continuity.

Self-Hosting & Configuration

• Install the Python package via pip and the Blender addon from the included folder
• Requires Blender 3.0 or later with Python scripting enabled
• Server runs on a configurable local port, defaulting to localhost
• Compatible with Claude Desktop, Cursor, and other MCP-capable clients
• Add the server to your MCP client config with the blender-mcp command

Key Features

• Natural language 3D scene creation without learning Blender's complex UI
• Real-time scene inspection that gives AI full context of the 3D workspace
• Support for poly mesh editing, primitive creation, and boolean operations
• Material and texture management through descriptive natural language prompts
• Script execution tool for unlimited extensibility beyond built-in operations

Comparison with Similar Tools

• Blender Python API — Direct scripting requires Python expertise; Blender MCP abstracts this behind natural language
• Three.js — Web-based 3D; Blender MCP leverages Blender's full production-grade toolset
• ComfyUI — Focused on 2D image generation; Blender MCP operates in true 3D space
• OpenSCAD — Programmatic CAD; Blender MCP handles organic modeling and rendering that OpenSCAD cannot

FAQ

Q: Which AI assistants work with Blender MCP? A: Any MCP-compatible client works, including Claude Desktop, Cursor, and custom MCP clients.

Q: Can the AI render images from the scene it creates? A: Yes, there is a render tool that triggers Blender's renderer and returns the output image.

Q: Does it work with Blender's geometry nodes? A: The script execution tool can manipulate geometry nodes, though there are no dedicated high-level tools for it yet.

Q: Is real-time collaboration supported? A: Blender MCP is single-user. One AI client connects to one Blender instance at a time.

Sources

• https://github.com/ahujasid/blender-mcp
• https://modelcontextprotocol.io