What is cc-blender-skill?
cc-blender-skill is a skill pack that connects an AI coding agent to Blender through an MCP backend, either the official Blender Lab MCP on Blender 5.1 or ahujasid/blender-mcp on 4.0. It chains 30 skills covering modeling, materials, lighting, cameras, rendering, animation and export, converting a plain request into Blender Python the agent runs for you.
cc-blender-skill is a skill pack that lets an agent drive Blender from a plain request, such as modeling a sword and rendering it with three-point lighting. It converts that request into Blender Python through one of two MCP backends.
It needs the official Blender Lab MCP on Blender 5.1 and later, or the community ahujasid/blender-mcp project on Blender 4.0 and later, plus Python 3.9 or newer with opencv-python, numpy, scipy and Pillow for image work.
What the 30 skills cover
- Core domains: modeling, materials, lighting, cameras, rendering and animation, each a separate chain-loadable skill rather than one giant prompt.
- Conversion work: wireframe-to-3D reconstruction and rebuilding a shape from a reference image.
- Output work: export to common 3D formats, plus UV and texture atlas fitting.
- Quality passes: validation, repair and optimization skills that check a model before it is treated as done.
What it does not try to fix
The maintainer states plainly that the skill pack produces functionally correct geometry, not finished art direction. Aesthetic judgment, human faces built from primitive shapes, and thin-metal specular detail still need a person's eye and a manual pass.
Who it is for
Best for
- Agent workflows that need a first-pass 3D asset, such as a prop or a simple scene, built from a short text description.
- Turning a wireframe sketch or a reference photo into a rough 3D starting point for a human to refine.
- Pipelines that need Blender objects validated and repaired automatically before export.
Not for
- If the asset lives inside a Figma Desktop file rather than a Blender scene, Conduit Design MCP can create, edit and export that instead.
- When the canvas to work from is a Figma design file instead of a Blender scene, Figma MCP Server reads that design context and can generate code from it.
- If what you need is an animated flow diagram of a system rather than a rendered 3D object, Dashmotion Skill builds that animated diagram instead.
- When the build is wiring a React or Python project's components together rather than modeling in 3D, assembling-components Skill wires those project pieces together.
Use it with UiChemy
cc-blender-skill builds and renders the 3D asset itself, outside any web page. Once that render is ready as an image or video, UiChemy Composer places it inside an editable WordPress page. Click Get Started on uichemy.com.
How to install cc-blender-skill: Model and Render in Blender by Request
git clone [email protected]:RobLe3/cc-blender-skill.gitBlender 5.1 or later is needed for the official MCP route; 4.0 is the floor for the ahujasid/blender-mcp route.Clone the repository
Run
git clone [email protected]:RobLe3/cc-blender-skill.gitthencd cc-blender-skill. A direct clone is the documented path, with no packaged installer command.Link the skills folder
Create
$HOME/.agents/skillsif it does not exist, then symlink every folder underplugin/skills/into it, so your agent host picks up all 30 skills at once.Set up an MCP backend
Install either the official Blender Lab MCP for Blender 5.1 and later, or ahujasid/blender-mcp for Blender 4.0 and later, and connect it to your agent as you would any other MCP server.
Ask for a build
Try a request like "Make a 3D model of a teapot and render it with three-point lighting" and let the chained skills carry out the modeling, lighting and render steps.
Check that it loaded
Ask the agent for a simple object and a render. It should walk through a modeling skill, a lighting skill and a rendering skill in sequence and hand back an image or a saved Blender scene.
Key features
- Natural-language to Blender PythonA plain request such as modeling a teapot and rendering it with three-point lighting gets converted into Blender Python the agent runs.
- 30 chain-loadable skillsModeling, materials, lighting, cameras, rendering, animation, export and more each load as their own skill, called in sequence as a task needs them.
- Two supported MCP backendsWorks with the official Blender Lab MCP on Blender 5.1 and later, or the community ahujasid/blender-mcp project on Blender 4.0 and later.
- Wireframe and reference reconstructionSeparate skills turn a wireframe sketch or a reference image into a rough 3D starting shape for further work.
- UV and atlas fittingHandles texture UV layout and atlas fitting as its own step, rather than leaving unwrapped geometry for a person to fix.
- Validation and repairDedicated skills check a model for problems and repair or optimize it before it is treated as finished.
- Honest scopeThe maintainer documents upfront that aesthetic polish, primitive-built human faces and thin-metal specular detail still need a person's pass.
Frequently asked questions
What is cc-blender-skill?
It is a skill pack that connects an AI coding agent to Blender through an MCP backend, converting a plain request such as modeling a sword into executable Blender Python across 30 chain-loadable skills covering modeling, lighting, rendering and more.
How do I install cc-blender-skill?
Clone the repository, then symlink every folder under plugin/skills into your agent's skills directory, for example $HOME/.agents/skills. A clone plus that symlink step is the documented path; there is no packaged installer.
Which Blender version does cc-blender-skill need?
Blender 5.1 or later for the official Blender Lab MCP route, or Blender 4.0 or later for the community ahujasid/blender-mcp route. Either backend needs Python 3.9 or newer with a short list of imaging packages.
Can it turn a reference image into a 3D model?
Yes, one of the 30 skills handles reference-image reconstruction, building a rough 3D starting shape from a photo or sketch. A wireframe-to-3D skill covers the related case of converting a wireframe drawing instead.
Does it finish the artwork for me?
No. The maintainer is explicit that the skills produce functionally correct geometry, not finished art direction. Aesthetic polish, primitive-built human faces and thin-metal specular detail still need a person's manual pass afterward.
Does it check a model before export?
Yes. Dedicated validation and repair skills check and fix a model, and a separate skill handles UV and texture atlas fitting, before the model is treated as ready for export.
Does cc-blender-skill work with agents other than Claude?
The README names Claude and Codex as compatible hosts, and says the skill pack works with any host able to load the plugin/skills folder and talk to the same MCP backend, so the format itself is not limited to one agent.


