blender-open-mcp
@dhakalnirajan
Open Models MCP for Blender Using Ollama
Overview
What is blender-open-mcp?
Control Blender 3D with natural language prompts via local AI models. It connects MCP clients (Claude, Cursor, or any MCP client) to Blender through a local Ollama LLM. Built on the Model Context Protocol.
How to use blender-open-mcp?
Install prerequisites (Blender 3.0+, Python 3.10+, Ollama, uv). Clone the repo, create a venv with uv, install the package, install the Blender add-on (addon.py), pull an Ollama model (e.g. llama3.2), start Ollama, then run blender-mcp (with optional host/port/model flags). For stdio transport (Claude Desktop, Cursor) use --transport stdio. Use the bundled CLI (blender-mcp-client) or the Python API to interact.
Key features of blender-open-mcp
- Natural language control of Blender via local AI (Ollama)
- Supports multiple MCP transports (HTTP and stdio)
- Full Blender scene inspection and object manipulation
- Material assignment, rendering, and PolyHaven asset integration
- Execute arbitrary Python/bpy code in Blender
- Switch active Ollama model and server URL at runtime
Use cases of blender-open-mcp
- Create and modify Blender scenes using plain English prompts
- Automate repetitive tasks (adding objects, applying materials, rendering)
- Integrate AI-assisted code generation into Blender pipelines
- Browse and import PolyHaven textures and materials
- Control Blender remotely from any MCP-compatible client
FAQ from blender-open-mcp
What are the prerequisites for blender-open-mcp?
Blender 3.0+, Python 3.10+, the latest Ollama, and uv. An internet connection is needed to pull Ollama models and PolyHaven assets.
How do I connect Claude Desktop or Cursor to blender-open-mcp?
Add the server configuration to your mcp.json (or ~/.cursor/mcp.json) with the command "blender-mcp" and args ["--transport", "stdio"].
Why can't the server connect to the Blender add-on?
Open Blender, go to the 3D Viewport N‑sidebar, find the Blender MCP panel, and click "Start MCP Server". The add‑on listens on port 9876 by default.
What ports does blender-open-mcp use?
FastMCP server: port 8000 (configurable). Blender add‑on TCP socket: port 9876. Ollama: port 11434.
Can I use a different Ollama model?
Yes. Pull any model with ollama pull <model> and specify it at server start with --ollama-model <model>, or switch it later via the blender_set_ollama_model tool.