MCP in Blender & Unity with Claude AI#

Recently, I experimented with MCP (Model Context Protocol) to connect 3D tools like Blender and Unity with Claude AI. My goal was straightforward: see if AI could directly manipulate or inspect my 3D environments in real time. This post is less of a polished how-to and more of a developer log — what worked, what broke, and what I learned.

Understanding MCP#

MCP (Model Context Protocol) is an open protocol that standardizes how applications talk to large language models (LLMs).

Key ideas I had to wrap my head around:

Two-way communication between AI and programs.
Ability to create, modify, or delete 3D objects.
Editing materials, inspecting entire scenes, and even executing code through the AI.

💡 Why this matters: Instead of AI being a “detached helper,” MCP turns it into an active collaborator inside your workflow. That was the hook for me.

Tools & Sources#

Blender 3.0+
Unity 6+
Python 3.10+ (Blender MCP), 3.12+ (Unity MCP)
Blender MCP GitHub
Unity MCP GitHub
Claude AI App
UV Package Manager

Step-by-Step Setup#

Setup#

1. UV Package Manager#

This is required for both Blender and Unity MCP.

Open terminal and install UV Package Manager.
If the first command failed, switch to the alternative one provided in the docs.

💡 Why: MCP depends on the package manager for dependency handling. Without this, nothing connects properly.

Blender MCP Setup#

1. Addon Installation#

addon

Download addon.py from Blender-MCP GitHub.
In Blender, go to Edit → Preferences → Add-ons.
Click Install from Disk and select the addon.py file.
Enable the addon by checking the box next to “Interface: Blender MCP”.

💡 Why: This gives Blender a sidebar tab for MCP. Without the addon, Blender doesn’t know how to “talk” to Claude.

2. Blender MCP Tab#

blender_mcp

Open the 3D View sidebar (press N).
Find the BlenderMCP tab.
Check Poly Haven if you want asset streaming.
Click Connect to Claude.

💡 Why: This is where you actually link Blender’s context to Claude. It felt surreal seeing AI recognize what was in my scene.

3. Claude AI Config#

claude-blender

Open Claude AI App and go Settings → Developer → Edit Config.
Copy and paste the below code in claude_desktop_config file.

1
{
2
  "mcpServers": {
3
    "blender": {
4
      "command": "uvx",
5
      "args": ["blender-mcp"]
6
    }
7
  }
8
}

Restart the computer.

💡 Why: This step gives Claude the bridge it needs to process MCP requests. Took me a few tries before it clicked.

Video#

Unity MCP Setup#

1. Python Path Setup (Windows)#

windowpath

Open Python file and click “Open File Location”.
Copy the file path.
Open “Edit the system environment variables”.
Click Environment Variables.
Click Path in System variables and Edit.
Paste the file path.

2. Package Installation#

packm

In Unity, open Window → Package Manager.
Click + → Add package from Git URL....
Paste the link: https://github.com/CoplayDev/unity-mcp.git?path=/UnityMcpBridge.
Click Add.

💡 Why: MCP for Unity isn’t built-in — GitHub package installation is the only way to get it.

3. MCP Auto-Setup#

mcp_auto

Go Window → Unity MCP.
Click Auto-Setup.
Look for a green status indicator🟢 and “Connected ✓”. (This attempts to modify the MCP Client’s config file automatically).
Restart the computer.

💡 Why: Auto-setup edits the MCP client config automatically. Saves time and avoids manually editing JSON files.

Video#

Discussion & Reflections#

Challenges: MCP requires detailed context to be useful. If you don’t feed enough scene info, the AI’s edits are random or incomplete.
What worked: Once connected, I could spawn and modify objects in Blender directly through Claude — faster than manual modeling tweaks.
Limitations: MCP helps speed up repetitive tasks. But currently it doesn’t replace a human designer at all. You still need to think critically about scene composition.

Future Outlook#

Game Development: AI-assisted real-time prototyping inside Unity.
Simulation: Rapid environment adjustments for training data.
VR/AR Content Creation: AI could manage assets or materials while you focus on immersion.
Labor Impact: Might reduce grunt work, but creative direction still belongs to humans.