Blender MCP is an open-source project that connects AI assistants to Blender 3D through the Model Context Protocol (MCP). It consists of a Blender addon running a socket server and an MCP server that relays commands from the AI to Blender. You describe what you want in natural language, and the AI translates it into Blender operations.

Which AI models work with Blender MCP?

All major ones: Claude Desktop, ChatGPT, Gemini, Cursor IDE, VS Code (via Cline/Roo Code), and local models via Ollama. Since it uses the standard MCP protocol, any AI client that supports MCP can potentially connect.

Yes. It is fully open-source under the MIT license. The GitHub repository has 24,400+ stars and has been updated recently.

Do I need to know Python or Blender's API?

No. The whole point is to let you describe scenes in plain English. Power users can also ask the AI to write and execute Python code in Blender, but it is not required for basic use.

Can it generate new 3D models from scratch?

Yes, through integrations with Hyper3D Rodin and Hunyuan3D. You can prompt the AI to add a new 3D object, and it will generate the geometry using these services. It can also fetch models from Sketchfab and Poly Haven.

Blender + AI: How Blender MCP Makes 3D Modeling as Easy as Describing What You Want

The One-Line Setup

Add this JSON entry to your claude_desktop_config.json:

{
  "mcpServers": {
    "blender": {
      "command": "uvx",
      "args": ["blender-mcp"]
    }
  }
}

That single block is the entire configuration. Install uv if you haven't already, download and install the addon.py in Blender's addon preferences, click Connect, and you are done.

This is what got my attention. Most AI-creative tool integrations take an afternoon of fiddling. This one took me under two minutes.

What It Can Do

Natural Language Modeling

"Create a low-poly dungeon with a dragon guarding a pile of gold." Blender MCP takes that sentence and builds the scene — objects, lighting, camera, materials. Every command goes through Blender's native operations, so you can tweak anything afterwards just like you would with manually created geometry.

Scene Inspection and Modification

Ask the AI what is in your current scene: "Describe this scene and tell me how many objects have materials assigned." Then modify it: "Select all wall objects and change their material to a dark stone texture."

Material and Lighting

Set moods with natural language: "Make this look like a sunrise with warm golden light coming from the right." The AI sets up studio lighting, applies materials, and adjusts camera angles.

AI 3D Model Generation

Through Hyper3D Rodin and Hunyuan3D integration, you can generate new 3D models from text prompts directly inside Blender: "Add a stylized palm tree next to the beach house."

Asset Library Search

Search and download models from Sketchfab, or pull HDRIs, textures, and 3D models from Poly Haven — all by describing what you want.

Python Code Execution

For advanced use: "Write a Python script that creates 100 cubes arranged in a Fibonacci spiral with random colors." The AI writes the script, executes it in Blender, and you see the result instantly.

Works with Every Major AI Client

| AI Client | Setup Time | Best For | |-----------|-----------|----------| | Claude Desktop | ~2 min | Most polished, full MCP support | | ChatGPT | ~3 min | Familiar interface | | Gemini | ~3 min | Google ecosystem integration | | Cursor / VS Code | ~2 min | Developers inside their IDE | | Ollama (local) | ~5 min | Free, offline, fully private |

Why This Matters

For Designers and 3D Artists

The 3D modeling process has a lot of mechanical repetition — placing primitives, adjusting transforms, setting up materials, arranging lights. Blender MCP handles the boilerplate so you can focus on creative decisions. Instead of clicking through five menus to create an array of pillars, you say: "Create a row of six columns, evenly spaced, with a Doric style."

For Architects and Visualizers

Block out an entire building with a single sentence: "Create a modern two-story house with a flat roof, large glass windows on the south facade, and a wooden deck on the east side." Iterate through conversation. The architecture community has already published impressive workflow demos.

For Game Developers

Rapid prototyping of environments, props, and level layouts. The Blender-to-Unreal/Unity pipeline becomes dramatically faster when you can describe the scene once instead of building it piece by piece.

For Anyone Curious About AI + Creativity

Blender MCP is one of the best real-world demonstrations of how the Model Context Protocol can reshape creative software. Instead of every tool building its own "AI button," MCP gives us a universal connector. The same pattern that connects AI to Blender could connect to Photoshop, Unreal Engine, or any application that exposes an API.

The Architecture (Simple, But Clever)

Blender MCP works in three parts:

Blender Addon (addon.py) — runs a socket server inside your Blender instance. Listens for commands and executes them against Blender's Python API.
MCP Server (the blender-mcp package) — runs as a subprocess alongside your AI client. Translates MCP tool calls into socket messages for the addon.
AI Client — Claude, ChatGPT, or any MCP-compatible app. You type prompts, the client sends tool requests to the MCP server, and commands flow into Blender.

All communication stays on your local machine. No data leaves your network unless you use cloud-based AI models (and even then, the Blender scene data stays local — only the text prompt goes to the model).

Getting Started (The Quick Version)

Install uv — the fastest Python package manager
Install the Blender addon — download addon.py from GitHub, install it in Blender (Edit → Preferences → Add-ons → Install)
Configure your AI client — paste the MCP server JSON (one-liner above)
Open Blender — start the addon by searching "MCP" in Blender's addon menu and clicking Connect (you will see "Connected" in the addon panel)
Start prompting — try something simple like "Create a red cube on a blue plane with a spotlight"

The setup is the same whether you use Claude, ChatGPT, Gemini, or a local model. The only difference is where you put the MCP config.

What Makes This Different From Other AI-3D Tools

There are other ways to generate 3D content with AI — text-to-3D generators, NeRF converters, image-to-mesh pipelines. Blender MCP takes a different approach:

It works inside Blender, not as a separate service. Everything it creates is immediately editable in the native Blender environment.
It is bidirectional. The AI can inspect the scene and make targeted changes, not just generate something from scratch.
It is modular. You can use any AI model, swap between them, or run locally.
It is open source. The MIT license means no vendor lock-in, no surprise pricing changes.

What I Would Like to See Next

Blender MCP is already impressive, but a few things would make it even better:

Animation support — the current tools focus on static scenes. Adding keyframe and animation rigging would unlock character work and motion design.
GPU-accelerated rendering — being able to say "render this scene in Cycles at 4K" and get the result back would close the loop nicely.
Multi-scene orchestration — managing multiple Blender instances or scenes through one AI session would be powerful for large projects.

The project is moving fast (the repo was updated just days ago), so I expect these features are on the roadmap.

Final Thoughts

Blender MCP is one of those tools that makes you feel like you are using the future. Two minutes of setup, a single sentence, and Blender builds a 3D scene. It is not perfect — complex geometry still needs manual cleanup, and the AI occasionally misunderstands spatial relationships. But the direction is unmistakable.

If you work with 3D in any capacity, set this up. It costs nothing, takes almost no time, and it will change how you think about the relationship between AI and creative tools.

Have you tried Blender MCP? Drop me a message on social media — I would love to see what you built with it.

Resources

Blender MCP on GitHub — 24.4k ★, MIT license
Official website — Setup guides for every AI client
Blender MCP Discord — Community, showcase, support
Blender MCP YouTube Tutorial — Full walkthrough