Devotd Github
NewModel Context Protocol

licincan/blender-ai-mcp-addon v1.0.1 — Scene Editing AI and a Built-in MCP Server

licincan/blender-ai-mcp-addon v1.0.1 provides an AI assistant that inspects Blender scenes and supports object creation, material editing, and asset search, along with a built-in MCP server.

Repositorylicincan/blender-ai-mcp-addon ↗

licincan/blender-ai-mcp-addon v1.0.1 provides an AI assistant that inspects Blender scenes and supports object creation, material editing, and asset search, along with a built-in MCP server.

GitHub description: AI assistant and MCP server for Blender: visual chat, scene tools, multi-agent workflows. Works with OpenAI, ChatGPT, Claude, Gemini, Ollama, OpenCode. Blender AI addon and plugin, copilot for 3D, bpy Python scripting, MCP integration.

stars
★ 959
Forks
389
Language
Python
License
MIT
Latest release
v1.0.1
Read original source ↗View on GitHub ↗

licincan/blender-ai-mcp-addon is both an add-on that lets you use an AI assistant inside Blender and an MCP server. The latest public release is v1.0.1, and the project is written in Python and released under the MIT license.

This add-on provides an AI assistant that inspects the Blender scene and performs object creation, material editing, asset search, and tasks spanning multiple steps. It ships with a visual chat workspace for handling the scene on screen and an MCP server embedded in the add-on, so you can either work with a single assistant or organize a team of Manager, Workers, and Reviewers. The README describes 21 scene tools and a procedure for running tasks in order and then reviewing them. External compatible clients can connect through authenticated Streamable HTTP MCP, and it is meant to be used with OpenAI, ChatGPT, Claude, Gemini, Ollama, and OpenCode. The repository has 959 GitHub stars.

However, the confirmed working environment is limited to Windows x64 and Blender 4.2 or later, and no performance comparison figures or actual adoption cases are presented. There is no information available on what changed in v1.0.1. The evaluation scores are 6.0 for freshness, 5.0 for technical merit, 4.0 for quality, and 7.6 for momentum, with quality coming out the lowest. It is worth considering as a way to attach AI to a 3D workflow, but what it actually supports, and how reliably, cannot be confirmed from the available material.