blender-llm-mcp

# blender-llm-mcp MCP (Model Context Protocol) server for Blender 3D — expose Blender modeling tools to AI assistants via the Model Context Protocol. GitHub: [WeitaoYan/blender-llm-assistant](https://github.com/WeitaoYan/blender-llm-assistant) ## Overview `blender-llm-mcp` is the **MCP server** component of the Blender LLM Assistant project. It acts as a bridge between AI assistants (Claude Desktop, VS Code, etc.) and the Blender 3D software. **This package alone is NOT enough** — it requires the [Blender LLM Assistant](https://github.com/WeitaoYan/blender-llm-assistant) addon to be installed and running inside Blender. The addon exposes an HTTP API on `127.0.0.1:15800`, and this MCP server translates MCP tool calls into HTTP requests to that API. ### Architecture ``` AI Assistant (Claude / VS Code) │ MCP stdio protocol ▼ blender-llm-mcp (this package) │ HTTP + SSE, Bearer Token auth ▼ Blender LLM Assistant addon ← must be running in Blender │ Blender Python API (bpy) ▼ Blender 3D scene ``` ## Usage Requires the [Blender LLM Assistant](https://github.com/WeitaoYan/blender-llm-assistant) addon running in Blender (HTTP API on `127.0.0.1:15800`). The Blender addon uses **Bearer Token authentication**. After starting the server in Blender, copy the token from the panel and pass it to the MCP server. ```bash # Install pip install blender-llm-mcp # Run with token blender-llm-mcp --blender-token <token-from-blender-panel> # Or with uvx (no install needed) uvx blender-llm-mcp --blender-token <token-from-blender-panel> # Custom Blender URL blender-llm-mcp --blender-url http://127.0.0.1:15800 --blender-token <token> ``` ## Configuration in Claude Desktop First set the token as an environment variable, then reference it in the config: ```bash export BLENDER_TOKEN="<token-from-blender-panel>" ``` Add to your `claude_desktop_config.json`: ```json { "mcpServers": { "bl…