npx mcpbar@latest install cwahlfeldt/blender-mcp -c claude

npx mcpbar@latest edit -c claude

{
  "mcpServers": {
    "Blender Script Manager": {
      "command": "uvx",
      "args": [
        "blender-mcp"
      ],
      "env": {}
    }
  }
}

Blender MCP Server

A Model Context Protocol (MCP) server for managing and executing Blender scripts.

Features

• Add, edit, execute, and remove Blender Python scripts
• Execute scripts in a headless Blender environment
• View execution results and errors
• Track script metadata (creation date, last modified, execution count)

Requirements

• Python 3.7+
• Blender installed and accessible
• MCP library (pip install mcp)

Usage

1. Start the server:

   python server.py
   

2. Connect to the server using an MCP client (like Claude Desktop)

3. Use the provided tools to manage scripts:

   • add_script(name, content) - Add a new script
   • edit_script(name, content) - Edit an existing script
   • execute_script(name, blend_file=None) - Execute a script in Blender, optionally specifying a .blend file
   • remove_script(name) - Remove a script

4. Access resources to get information:

   • scripts://list - Get list of available scripts
   • script://{name} - Get content of a specific script
   • result://{name} - Get execution result of a script

Examples

Basic Example

# Add a simple script
add_script("hello_cube", '''
import bpy

# Clear existing objects
bpy.ops.object.select_all(action='SELECT')
bpy.ops.object.delete()

# Create a cube
bpy.ops.mesh.primitive_cube_add(size=2, location=(0, 0, 0))
print("Cube created!")
''')

# Execute the script
execute_script("hello_cube")

# Get the result
# Access using: result://hello_cube

Working with Blend Files

# Add a script that works with a blend file
add_script("analyze_scene", '''
import bpy

# Print information about the current scene
print(f"Current Blender version: {bpy.app.version_string}")
print(f"Current file: {bpy.data.filepath}")

# List all objects in the scene
print("\\nObjects in the scene:")
for obj in bpy.data.objects:
    print(f" - {obj.name} ({obj.type})")
''')

# Execute with a specific blend file
execute_script("analyze_scene", blend_file="/path/to/your/project.blend")

# Get the result
# Access using: result://analyze_scene

How It Works

1. When a script is added, it's stored in the script_files/scripts directory
2. When executed, the script is run in a headless Blender instance
   • If a blend file is specified, Blender will open that file before running the script
   • Otherwise, a default empty Blender scene is used
3. Output and errors are captured and stored in the script_files/results directory
4. Metadata about scripts is tracked in script_files/metadata.json

Installation

1. Clone this repository
2. Install the MCP library: pip install mcp
3. Ensure Blender is installed and accessible from your PATH

License

MIT