YOU COMMENTED "BLENDER" — HERE IT IS

5 AI tools that let Claude build, rig and animate 3D

By ·

Five free, open tools that turn an AI agent into a 3D assistant: build scenes in Blender from plain English, follow real Blender workflows, turn one photo into a textured 3D model, rig a character, and loop until the render looks right. Each one was checked against its own repository on 16 Nov 2026, with the catches listed honestly.

1. MCP for Blender (formerly blender-mcp)

Repo: github.com/ahujasid/mcp-for-blender ↗ (MIT, about 34.6k stars)

What it does

Connects Claude (or any MCP client) to Blender, so the agent can create, modify and delete objects, set materials, check renders and run Blender Python.

Setup

one-command setup
uvx mcp-for-blender setup
or in Claude Code
claude mcp add blender uvx mcp-for-blender
then the add-on
uvx mcp-for-blender install-addon

Then open Blender (3.0+, Python 3.10+).

2. Blender Skills (EL4CTEO/blender-skills)

Repo: github.com/EL4CTEO/blender-skills ↗ (MIT)

What it does

28 agent skills that cover the 3D pipeline, from modeling, UVs and PBR materials to rigging, animation, export and rendering, so the agent follows tested workflows instead of guessing.

Setup

In Claude Code, run these two lines one after the other.

claude code · 1 of 2
/plugin marketplace add EL4CTEO/blender-skills
claude code · 2 of 2
/plugin install blender-agent-skills@blender-agent-skills

Other agents: clone the repo and run install/install.sh.

3. Hunyuan3D 2.1 (Tencent)

Repo: github.com/Tencent-Hunyuan/Hunyuan3D-2.1 ↗
Free demo: huggingface.co/spaces/tencent/Hunyuan3D-2.1 ↗

What it does

One image in, a 3D mesh with PBR textures out (open weights).

4. UniRig (VAST AI Research)

Repo: github.com/VAST-AI-Research/UniRig ↗ (MIT, SIGGRAPH 2025)

What it does

Predicts a skeleton and skinning weights for a 3D model (obj, fbx, glb, vrm), so it is ready to animate.

Setup

Python 3.11 environment, then install the requirements.

terminal
pip install -r requirements.txt

Then run generate_skeleton.sh, generate_skin.sh and merge.sh.

5. dream-loop

Repo: github.com/achimala/dream-loop ↗ (MIT, by @anshuc)

What it does

An agent skill that generates a target image, builds toward it, and has a separate critic agent compare screenshots and send fixes back, loop after loop.

Setup

terminal
npx skills add achimala/dream-loop

Or clone it into your skills folder.

Try this prompt first

Once MCP for Blender is connected, paste this into Claude and watch the scene appear in Blender.

Snowy cabin scene in Blender

prompt
Using the Blender MCP tools, build a snowy cabin at night with a warm lamp in the window, two pine trees and soft blue moonlight. Keep the scene under 50k polygons, add simple PBR materials, then render a preview and tell me what you would improve.

The order to try them in

  1. Install MCP for Blender (uvx mcp-for-blender setup) and open Blender with safe mode on
  2. Add Blender Skills to Claude Code with /plugin marketplace add EL4CTEO/blender-skills
  3. Prompt Claude to build a scene and check the render
  4. Make a 3D model from a photo in the Hunyuan3D 2.1 Hugging Face demo
  5. Rig it with UniRig on a CUDA GPU, then animate it in Blender
  6. Add dream-loop (npx skills add achimala/dream-loop) to let a critic agent refine the result

Sources checked

Checked on 16 Nov 2026. Star counts move every day and projects change their install steps, so each repo's own README is the final word.

  • MCP for Blender: MIT, about 34.6k stars, formerly blender-mcp, runs Blender Python, safe mode and telemetry flags. Source ↗
  • Blender Skills: 28 skills, MIT, plugin install lines, Blender 5.2 LTS+. Source ↗
  • Hunyuan3D 2.1: image to textured mesh, VRAM needs, community license territory limits. Source ↗
  • Hunyuan3D 2.1 hosted demo. Source ↗
  • UniRig: skeleton and skinning prediction, MIT, SIGGRAPH 2025, CUDA GPU. Source ↗
  • dream-loop: critic loop, MIT, by @anshuc, install command, stops after 3 loops by default. Source ↗

About the video: the reel that sent you here is presented by an AI avatar of Hammad Yousuf.

Keep going