Home / MCP / Blender MCP Server

Blender MCP Server

Controls Blender with local AI models via MCP, enabling natural-language prompts for modeling tasks and renders.

python

61stars

View on GitHub

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
    "mcpServers": {
        "blender_http": {
            "url": "http://0.0.0.0:8000"
        }
    }
}

You can drive Blender with local AI models through MCP, enabling natural-language prompts to perform 3D tasks, manage scenes, and render outputs. This setup lets you run a local Ollama model, connect it to Blender via a dedicated MCP server, and use a Blender add-on to communicate seamlessly with the AI-driven workflow.

How to use

Use an MCP client to send prompts that control Blender and query scene information. You can create objects, modify materials, render images, and download PolyHaven assets directly through prompts or tool actions. Start the MCP server, then run the client commands to issue actions such as retrieving scene details, creating objects, applying materials, and triggering renders.

How to install

Prerequisites to get started:

Blender 3.0 or later from blender.org
Ollama from ollama.com with suitable model installed
Python 3.10 or later
uv package (install via pip: pip install uv)
Git to clone the project

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

Create and activate a Python virtual environment using uv and install the package in editable mode.

uv venv
source .venv/bin/activate  # Linux/macOS
.venv\Scripts\activate  # Windows

uv pip install -e .

Install the Blender add-on by loading the provided addon file into Blender. In Blender, enable the add-on in the Preferences.

# Steps inside Blender UI
# Edit -> Preferences -> Add-ons
# Install... and select addon.py from blender-open-mcp directory
# Enable the Blender MCP add-on

Prepare an Ollama model in advance. If you need to pull a model, you can run a command to ensure the model is available.

ollama run llama3.2  # or another supported model like Gemma3

Setup

Start the Ollama server in the background, then start the MCP server and finally enable communication with Blender through the add-on.

Start the MCP server (HTTP endpoint by default at port 8000):

blender-mcp
# or with explicit host/port and Ollama settings
blender-mcp --host 127.0.0.1 --port 8001 --ollama-url http://localhost:11434 --ollama-model llama3.2

Alternatively, run the Python-based server directly:

python src/blender_open_mcp/server.py

Start the Blender add-on server

Open Blender, access the 3D Viewport, and start the integrated MCP server from the Blender MCP panel.

# In Blender UI: N key to open the sidebar, locate the Blender MCP panel, and click Start MCP Server

Usage with the MCP client

Interact with the server using the mcp client to issue prompts and tool commands. You can perform basic operations, query scene data, and render outputs. Use prompts to create objects, apply materials, or trigger renders, and rely on the tool list to perform more advanced actions.

mcp prompt "Create a cube named 'my_cube'." --host http://localhost:8000
mcp tool get_scene_info --host http://localhost:8000
mcp prompt "Render the image." --host http://localhost:8000

PolyHaven integration

Optionally, download PolyHaven assets through prompts when enabled in MCP. You can request textures, HDRIs, or models directly into your Blender scene.

mcp prompt "Download a texture from PolyHaven." --host http://localhost:8000

Troubleshooting

If you encounter issues, verify that Ollama and the MCP server are running, check the Blender add-on settings, review the startup arguments, and inspect logs for error messages.

Available tools

get_scene_info

Retrieves details about the current scene, including objects, camera, and render settings.

get_object_info

Retrieves information about a specific object by name.

create_object

Creates a 3D object with type, name, location, rotation, and scale.

modify_object

Modifies an object's properties such as location, rotation, scale, and visibility.

delete_object

Deletes an object by name.

set_material

Assigns a material to an object, including color customization.

render_image

Renders an image to a file path.

execute_blender_code

Executes Python code inside Blender to perform custom actions.

get_polyhaven_categories

Lists PolyHaven asset categories available for browsing.

search_polyhaven_assets

Searches PolyHaven assets by type and categories.

download_polyhaven_asset

Downloads a PolyHaven asset with specified type and resolution.

set_texture

Applies a downloaded texture to a specified object.

set_ollama_model

Sets the Ollama model to use for prompts.

set_ollama_url

Sets the Ollama server URL used for prompts.

get_ollama_models

Lists available Ollama models.