BlenderMCP: Control Blender 3D With AI Using Natural Language (2026 Setup Guide)

Blender is the world's most popular open source 3D creation suite, but its interface is notoriously complex. Creating a simple scene can mean digging through dozens of menus, writing Python scripts, and reading API documentation for hours. For beginners and experienced artists alike, the friction is real.

BlenderMCP, created by developer Sid Ahuja (ahujasid), is a free, open source Model Context Protocol (MCP) server that bridges AI coding agents (like Claude Desktop or Cursor) directly to Blender. Instead of manually writing Python scripts and pasting them into Blender's text editor, you simply describe what you want in natural language and watch it appear in your 3D viewport in real time.

In this guide, you'll learn what BlenderMCP is, who it's for, and how to set it up in under five minutes.

What is BlenderMCP?

BlenderMCP is an open source Model Context Protocol (MCP) server that acts as a two-way communication bridge between AI coding agents and Blender. Historically, if you wanted an AI to help you build a 3D scene, it would generate raw Python scripts that you had to manually copy and paste into Blender's text editor, often resulting in broken code due to API changes. BlenderMCP fixes this by exposing a set of reliable, predefined tools that allow the AI to natively interact with your active 3D scene in real time.

By running a lightweight socket server directly inside Blender, your AI assistant can analyze your currently selected objects, read your modifier stacks, spawn new meshes, assign colors, and even automatically fetch and apply professional assets from external libraries like Poly Haven and Sketchfab, all based on your natural language conversational prompts.

Who is it for?

• 3D artists and animators looking to automate tedious workflows, such as renaming hundreds of data blocks or diagnosing why a heavy scene is rendering slowly.
• Game developers wanting to rapidly prototype block-out levels and environmental layouts using natural language before committing to detailed modeling.
• Absolute beginners who find Blender's notoriously complex user interface overwhelming and want an AI co-pilot to translate their ideas directly into the 3D viewport.
• Python developers who want a safe, isolated pipeline to execute and test generative 3D scripts on the fly.

What makes BlenderMCP different from writing Python scripts manually?

• Direct viewport control: It doesn't just write code; it executes it. You can tell Claude to "create a procedural brick wall," and you will watch the bricks appear and stack dynamically in your Blender window. No copy-pasting required.
• Asset library integration: It connects the AI directly to external APIs. You can prompt the agent to "download a wooden texture from Poly Haven and apply it to the floor," and the MCP server will handle the fetching and node setup autonomously.
• Scene inspection and telemetry: The AI can actually see your project data. It can scan your scene for performance bottlenecks (like an overly subdivided mesh) and even capture viewport screenshots to visually understand the spatial composition of your models.
• Hyper3D Rodin support: It includes integrations for AI 3D model generation, allowing you to prompt a 2D image or text and have the resulting 3D geometry imported directly into your scene.
• Free and open source: The entire bridge is free and open source. You only pay for your AI client's API usage (Claude, etc.). No enterprise licensing, no monthly fees.

What you need before you start

To run this bridge effectively, you need to prepare both your 3D software and your AI environment:

• Blender 3.0 or newer: Ensure your 3D software is up to date.
• Python 3.10+: Required for running the background server.
• The uv package manager: An ultra-fast Python package installer required to run the server environment.
• An MCP client: You must have an AI application that supports the Model Context Protocol, such as the Claude Desktop app, Cursor, or Intel AI Playground.

Step-by-step installation

Step 1: Install the uv package manager

If you don't have it already, install uv to manage the Python server. Open your terminal or PowerShell:

Mac/Linux:

curl -LsSf https://astral.sh/uv/install.sh | sh

Windows:

irm https://astral.sh/uv/install.ps1 | iex

Step 2: Download the repository

Clone the BlenderMCP project to a permanent location on your hard drive:

git clone https://github.com/ahujasid/blender-mcp.git
cd blender-mcp

Step 3: Install the Blender add-on

The system works in two parts. First, you must install the listener inside Blender:

1. Inside the downloaded blender-mcp folder, locate the addon.py file.
2. Open Blender, navigate to Edit > Preferences > Add-ons.
3. Click Install and select the addon.py file.
4. Check the box to enable the add-on. In the 3D Viewport sidebar (press N), you will now see a BlenderMCP tab.

Step 4: Connect Claude Desktop

You now need to tell your AI client where the server is located. Open your Claude Desktop configuration file:

Mac: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

Add the following configuration, ensuring you replace the directory with your actual absolute path:

{
  "mcpServers": {
    "blender": {
      "command": "uv",
      "args": [
        "--directory",
        "/absolute/path/to/blender-mcp",
        "run",
        "blender-mcp"
      ]
    }
  }
}

Save the file and restart Claude Desktop. When you chat with Claude, you will see a hammer icon indicating the Blender tools are active. Ask it to "create a grid of 10 red spheres," and watch Blender go to work.

Common errors and how to fix them

Error	What it means	How to fix it
"Connection Refused" when Claude tries to use a tool	The MCP server is running, but the internal Blender socket is closed, so Claude has nowhere to send the commands.	Open Blender, press N to open the sidebar, go to the BlenderMCP tab, and click the Start Server button to begin listening for the AI.
Claude writes Python code in chat but doesn't execute it	The AI model misunderstood your intent and generated a markdown code block for you to copy, rather than actively calling the execution tool.	Reply to the AI with explicit instructions: "Do not give me the script. Use your Blender MCP tools to execute this code directly in the scene."
Path parsing errors on Windows	The claude_desktop_config.json file contains improperly formatted backslashes in the directory path.	JSON requires escaped backslashes. Ensure your path looks like C:\\Users\\Name\\blender-mcp instead of C:\Users\Name\blender-mcp.

BlenderMCP vs proprietary 3D AI assistants

Feature	BlenderMCP (open source)	Proprietary assistants (Promethean AI)
Cost	$0 (only pay your own AI API tokens)	Enterprise licensing or heavy monthly fees
Software target	Native Blender integration	Often locked to Maya or Unreal Engine
Extensibility	Open protocol (use Claude, ChatGPT, or local LLMs)	Closed ecosystem
Setup effort	Requires manual config files	Automated installers
Best for	Blender users who want free AI control of their viewport	Enterprise studios with dedicated tech art budgets

Bottom line: BlenderMCP completely changes how you interact with 3D software. Instead of digging through endless nested menus or reading Python API docs for hours, you can simply converse with your viewport. If you are already using an MCP-capable client like Claude Desktop, setting up this bridge is free and takes less than five minutes. It is a must-have for modern 3D workflows, especially for Blender users who want to leverage AI without paying enterprise licensing fees.

3 alternatives worth checking out

• Official Blender MCP Server (blender.org/lab/mcp-server): Recently launched by the official Blender development team via the Blender Lab. It shares a very similar goal of providing a natural language interface to Blender's Python API, with an explicit focus on scene analysis and debugging complex modifiers. It is slightly more official but may have different tool exposures than community-built servers.
• blender-ai-mcp (github.com/patrykiti/blender-ai-mcp): A more production-shaped MCP server. Instead of just exposing raw Python execution (which can be unstable), this repository focuses on bounded, verified workflow tools. It uses goal-first orchestration and assertions to ensure the AI doesn't break your scene, making it safer for professional pipelines.
• djeada/blender-mcp-server (github.com/djeada/blender-mcp-server): Another excellent community-built implementation. It exposes 27 highly specific tools across 7 namespaces, heavily categorized for creating objects, rendering images, exporting scenes, and managing asynchronous jobs without freezing the Blender viewport.

---

Found this guide useful? Check out more open source projects and free tools for developers on Sudo Scout.