Use the Blender skill to:
1. Check PolyHaven status
2. Download an HDRI for environment lighting
3. Download a wood texture
4. Create a cube and apply the wood texture
5. Render the scene
示例 2：导入 Sketchfab 模型
Search Sketchfab for "vintage car" and import the first downloadable result.
Scale it to fit the scene and position at origin.
示例 3：使用 Hyper3D 生成 AI 模型
Generate a 3D model of "a small wooden treasure chest with gold trim"
using Hyper3D Rodin. Import it and add environment lighting.
示例 4：自定义 Python 脚本
# Create a grid of cubes with random colors
import bpy
import random
for x in range(-5, 6):
for y in range(-5, 6):
bpy.ops.mesh.primitive_cube_add(location=(x*2, y*2, 0))
obj = bpy.context.active_object
mat = bpy.data.materials.new(name=f"Mat_{x}_{y}")
mat.diffuse_color = (random.random(), random.random(), random.random(), 1)
obj.data.materials.append(mat)
This skill enables Claude to interact with Blender for 3D modeling, scene manipulation, material application, and rendering operations through the BlenderMCP addon (ahujasid/blender-mcp - 14k+ stars).
Features
Core Capabilities
Two-way communication : Connect Claude AI to Blender through socket server
Object manipulation : Create, modify, and delete 3D objects
Material control : Apply and modify PBR materials and colors
Scene inspection : Get detailed information about the current scene
Code execution : Run arbitrary Python code in Blender
Viewport screenshots : Capture visual feedback from Blender
Asset Integrations
PolyHaven : Download HDRIs, textures, and 3D models
Sketchfab : Search and import models from the Sketchfab library
Hyper3D Rodin : Generate 3D models from text/images using AI
Hunyuan3D : Alternative AI 3D generation from Tencent
Prerequisites
Blender 3.0+ (recommended 4.0+)
BlenderMCP addon installed and active
Socket server running on port 9876 (default)
Docker Installation (Auto-configured)
The addon is automatically installed when building with Dockerfile.unified:
downloadable (optional): Only downloadable models (default: true)
download_sketchfab_model
Download and import a Sketchfab model.
uid (required): Sketchfab model UID
Hyper3D Rodin Integration
get_hyper3d_status
Check if Hyper3D Rodin is enabled.
generate_hyper3d_model_via_text
Generate 3D model from text description.
text_prompt (required): Description in English
bbox_condition (optional): [Length, Width, Height] ratio
generate_hyper3d_model_via_images
Generate 3D model from reference images.
input_image_paths (optional): List of image file paths
input_image_urls (optional): List of image URLs
bbox_condition (optional): Size ratio
poll_rodin_job_status
Check generation task status.
subscription_key: For MAIN_SITE mode
request_id: For FAL_AI mode
import_generated_asset
Import completed Hyper3D model.
name (required): Object name in scene
task_uuid: For MAIN_SITE mode
request_id: For FAL_AI mode
Hunyuan3D Integration
get_hunyuan3d_status
Check if Hunyuan3D is enabled.
generate_hunyuan3d_model
Generate 3D model using Hunyuan3D.
text_prompt (optional): Text description
input_image_url (optional): Reference image URL
poll_hunyuan_job_status
Check Hunyuan3D task status.
job_id (required): Job identifier
import_generated_asset_hunyuan
Import completed Hunyuan3D model.
name (required): Object name
zip_file_url (required): Generated model URL
Examples
Example 1: Create Scene with PolyHaven Assets
Use the Blender skill to:
1. Check PolyHaven status
2. Download an HDRI for environment lighting
3. Download a wood texture
4. Create a cube and apply the wood texture
5. Render the scene
Example 2: Import Sketchfab Model
Search Sketchfab for "vintage car" and import the first downloadable result.
Scale it to fit the scene and position at origin.
Example 3: Generate AI Model with Hyper3D
Generate a 3D model of "a small wooden treasure chest with gold trim"
using Hyper3D Rodin. Import it and add environment lighting.
Example 4: Custom Python Scripting
# Create a grid of cubes with random colors
import bpy
import random
for x in range(-5, 6):
for y in range(-5, 6):
bpy.ops.mesh.primitive_cube_add(location=(x*2, y*2, 0))
obj = bpy.context.active_object
mat = bpy.data.materials.new(name=f"Mat_{x}_{y}")
mat.diffuse_color = (random.random(), random.random(), random.random(), 1)
obj.data.materials.append(mat)
Standalone Server (Headless/Background)
For automated pipelines, use the standalone server instead of the UI addon:
# Start with VNC display (keeps Blender window open)
DISPLAY=:1 blender --python scripts/standalone_server.py
# Or with direct Python call
python3 -c "
import socket
import json
s = socket.socket()
s.connect(('localhost', 9876))
# Import a GLB mesh
request = json.dumps({
'type': 'import_model',
'params': {'filepath': '/path/to/mesh.glb', 'name': 'MyModel'}
})
s.sendall(request.encode())
print(s.recv(4096).decode())
s.close()
"
Standalone Server Commands
Command
Description
get_scene_info
Get scene details and object list
get_object_info
Get specific object properties
execute_blender_code
Run arbitrary Python code
import_model
Import GLB/OBJ/FBX/STL/PLY files
render
Render to image file
orbit_render
Render from multiple orbit angles
set_camera
Position camera at location/target
add_hdri
Add HDRI environment lighting
Integration with Other Skills
Works well with:
comfyui skill for text-to-3D model generation and validation
filesystem skill for managing output files
imagemagick skill for post-processing renders
ComfyUI 3D Integration (FLUX2 → SAM3D → Blender)
The Blender skill is the final validation step in the ComfyUI text-to-3D pipeline:
For visibility in the ComfyUI web interface, render directly to the ComfyUI output directory:
# If BlenderMCP is inside Docker with shared volumes:
output_dir = "/root/ComfyUI/output/validation"
# If Blender is on host, copy after rendering:
import shutil
import subprocess
# Copy from local to ComfyUI container
subprocess.run([
"docker", "cp",
"/tmp/validation/.",
"comfyui:/root/ComfyUI/output/validation/"
])
Or use docker exec to copy within the container network:
# Copy renders to ComfyUI output (from host)
docker cp /tmp/validation/. comfyui:/root/ComfyUI/output/validation/
# Or if both are containers in same network:
docker exec comfyui mkdir -p /root/ComfyUI/output/validation
docker cp blender:/tmp/renders/. comfyui:/root/ComfyUI/output/validation/
Performance Notes
Object creation: < 100ms
Material application: < 50ms
Asset downloads: varies by size (seconds to minutes)
AI generation (Hyper3D/Hunyuan): 30s - 5min depending on complexity
