blender-ai-mcp

blender-ai-mcp is a local Model Context Protocol server plus Blender addon bridge for AI-assisted 3D workflows.

It lets an MCP client inspect the live Blender scene and perform controlled operations on objects, materials, cameras, lights, render settings, animation, UVs, Geometry Nodes, imports, exports, and Python execution.

What Is Included

• server/: Python MCP server and tool registry.

• blender_addon/: Blender addon that receives commands from the local bridge.

• blender_ai_mcp/: package entry points for MCP clients.

• docs/: setup, tool reference, and AI usage notes.

Related MCP server: Bonsai MCP

Requirements

• Blender 3.6 LTS or newer.

• Python 3.10 or newer.

• Claude Desktop or another MCP-compatible client.

Quick Start

1. Create and activate a virtual environment.

   python -m venv .venv
   .\.venv\Scripts\Activate.ps1
   pip install -r requirements.txt

2. Install the addon from blender_addon/.

   In Blender, open Edit > Preferences > Add-ons > Install..., select blender_addon/__init__.py, and enable AI MCP Bridge.

3. Start the addon server in Blender.

   In the 3D Viewport sidebar, open the AI MCP panel, confirm port 9876, and click Start Server.

4. Configure your MCP client using mcp_config_example.json.

5. Restart the client and confirm the blender server is available.

MCP Config

Example:

{
  "mcpServers": {
    "blender": {
      "command": "python",
      "args": ["-m", "blender_ai_mcp.server.main"],
      "env": {
        "BLENDER_MCP_PORT": "9876",
        "BLENDER_MCP_HOST": "localhost"
      }
    }
  }
}

If you use a project virtual environment, set command to that environment's Python executable.

Typical Workflow

1. Inspect the live scene with get_scene_info and list_objects.

2. Create or reuse the target object.

3. Fix the origin if location and bounding_box_center do not match.

4. Position, align, and duplicate objects using the layout tools.

5. Apply materials and modifiers.

6. Validate overlaps, scale, and scene quality.

7. Render a viewport preview before reporting the task done.

Validation

Run the test suite after code changes:

pytest

Useful checks:

• ruff check .

• uv run --extra dev pytest tests

Documentation

• Setup guide

• Tool reference

• AI usage guide

Safety

The execute_python and expression-evaluation tools are for trusted local workflows only. Do not expose the Blender addon socket to a public network.

License

MIT. See LICENSE.