BlenderMCP
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@BlenderMCPCreate a red cube and a sphere, then render from camera"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
BlenderMCP — Blender Model Context Protocol Integration
BlenderMCP connects Blender to AI assistants through the Model Context Protocol (MCP), letting Claude directly control Blender for prompt-assisted 3D modelling, scene creation, rendering, and automation.
This is an extended fork of ahujasid/blender-mcp with 92 tools, full Blender 4.x compatibility, and telemetry removed.
What's new in this fork
92 tools across 26 categories (up from the original ~20)
Blender 4.x compatibility — fixes for BMesh layer API changes, EEVEE engine name,
temp_overrideregion requirement, compositor node orderingNo telemetry — all analytics code removed
Auto-restart on addon reload — server restarts automatically when the addon is cycled in Preferences; no manual click needed
deploy.py— one-command script to copyaddon.pyto Blender's installed addons pathimg_to_3d_server.py— local image-to-3D inference via TripoSR (optional, loads/unloads on demand to free VRAM)TOOLS.md— full reference for all 91 tools with parameters
Related MCP server: BlenderMCP
Tool categories (91 tools)
Category | Tools |
Process management |
|
Scene & object info |
|
Primitives & object mgmt |
|
Transforms |
|
Mesh editing |
|
Vertex operations |
|
Edge operations |
|
Face operations |
|
Curve control points |
|
Camera |
|
Lighting |
|
Materials |
|
Modifiers |
|
Rendering |
|
Viewport capture |
|
Reference images |
|
Animation |
|
Collections |
|
Export / import |
|
Scripting |
|
PolyHaven |
|
Sketchfab |
|
Hyper3D / Rodin |
|
Hunyuan3D |
|
Image-to-3D (TripoSR) |
|
See TOOLS.md for full parameter documentation.
Requirements
Blender 4.0 or newer (Blender 4.3 recommended)
Python 3.10+
uv package manager
Installation
1. Install the MCP server
git clone https://github.com/naab007/blender_mcp
cd blender_mcp
uv venv .venv
uv pip install -e .2. Install the Blender addon
Download
addon.pyfrom this repoOpen Blender → Edit → Preferences → Add-ons → Install...
Select
addon.pyand enable Interface: Blender MCPIn the 3D View sidebar (N), open the BlenderMCP tab and click Start MCP Server
After the first install, use
deploy.pyto push addon updates in one command:.venv/Scripts/python.exe deploy.pyEdit the destination path in
deploy.pyto match your Blender version.
3. Configure your AI client
Claude Code — create .mcp.json in your working directory:
{
"mcpServers": {
"blender": {
"type": "stdio",
"command": "/path/to/blender_mcp/.venv/Scripts/blender-mcp.exe"
}
}
}Claude Desktop — add to claude_desktop_config.json:
{
"mcpServers": {
"blender": {
"command": "uvx",
"args": ["blender-mcp"]
}
}
}Cursor — add to .cursor/mcp.json:
{
"mcpServers": {
"blender": {
"command": "uvx",
"args": ["blender-mcp"]
}
}
}Usage
Once the addon is running and the MCP server is configured, ask Claude to:
"Create a low-poly dungeon scene with a dragon guarding a pot of gold"
"Set up 3-point studio lighting and render from the active camera"
"Add a Subdivision Surface modifier to the selected object and apply a PBR material"
"Capture viewport angles from front, side, and top and give me a contact sheet"
"Boolean-subtract the sphere from the cube"
"Download a rock model from PolyHaven and place it at the origin"
"Generate a 3D model of a garden gnome via Hyper3D"
"Render depth map of the current scene"
"Keyframe this object moving from (0,0,0) to (5,0,0) over 60 frames"
"Compare these two renders and show me exactly what changed" →
diff_images(path_a, path_b)
Environment variables
Variable | Default | Description |
|
| Host for the Blender TCP socket |
|
| Port for the Blender TCP socket |
|
| Port for the local TripoSR server |
|
| Local weights path or HuggingFace hub ID |
| auto |
|
Dependencies
Package | Version | Purpose |
| ≥1.3.0 | MCP server framework |
| ≥10.0.0 | Image processing (viewport capture, contact sheets, diff) |
| ≥1.24.0 | Pixel math for |
| ≥3.0.0 | Local TripoSR inference server |
| ≥2.31.0 | PolyHaven / Sketchfab / Hyper3D API calls |
Architecture
Claude / AI client
│ MCP (stdio)
▼
server.py (FastMCP, @mcp.tool functions)
│ JSON over TCP :9876
▼
addon.py (Blender Python addon, bpy.app.timers dispatch)
│
▼
Blender sceneAll addon handlers run on Blender's main thread via
bpy.app.timers.register()— no threading issuesCommands are JSON
{ "type": "...", "params": {...} }, responses are{ "status": "success"|"error", "result": ... }
Troubleshooting
"context is incorrect" on view3d operators — requires Blender 4.x
temp_override(area=area, region=region). Already fixed in this fork.BLENDER_EEVEE enum not found — use
BLENDER_EEVEE_NEXTin Blender 4.x. Already normalized in this fork.Render writes nothing after depth map — compositor state corruption from empty node tree. Fixed in this fork via full snapshot/restore in
render_depth_map.execute_blender_coderuns arbitrary Python in Blender — powerful but irreversible. Save your work first.Connection issues — ensure the addon server is running (green dot in the BlenderMCP sidebar tab) before sending commands.
License
MIT — see LICENSE.
Based on ahujasid/blender-mcp by Siddharth Ahuja.
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Latest Blog Posts
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/naab007/blender_mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server