Skip to content
Checking statusStatus

Blender

Model, light and render in Blender by asking, through the Blender MCP add-on.

Model, light and render in Blender by asking: "Make a low-poly rocket, give it a metal body and render it." Claude or GPT does the work in your scene through the Blender MCP add-on, on your Evolved key.

Configure

  1. Connect an MCP app to evolved.to

    Blender is driven from an app that speaks MCP. Set one up first: Claude Code, VS Code or Cursor all work. See MCP servers for every app.

  2. Install uv

    The Blender MCP server runs with uvx, part of uv. Install it once; on a Mac with Homebrew that is brew install uv, and the uv site has the Windows installer.

  3. Add the add-on to Blender

    Download addon.py from the blender-mcp project. In Blender, open Edit → Preferences → Add-ons → Install from Disk, choose the file and tick Blender MCP.

  4. Start it in your scene

    In the 3D Viewport press N, open the BlenderMCP tab and click Connect. Blender now waits for your MCP app.

  5. Add the server to your app

    Add Blender to your app's MCP servers. For Claude Code, save this as .mcp.json in your project:

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

    Cursor, VS Code and the others take the same entry in their own file, listed on the MCP page. Restart the app and ask it to build something.

From a Python script

Blender's own Python can call evolved.to too, for example to name objects or write descriptions in an add-on of yours. Set EVOLVED_API_KEY in your environment before starting Blender, then run this in the Scripting workspace:

Text Editor · ask_evolved.py
import json
import os
import urllib.request

def ask_evolved(message):
    request = urllib.request.Request(
        "https://api.evolved.to/v1/chat/completions",
        data=json.dumps({
            "model": "gpt-5-mini",
            "messages": [{"role": "user", "content": message}],
            "max_tokens": 300,
        }).encode(),
        headers={
            "content-type": "application/json",
            "authorization": "Bearer " + os.environ["EVOLVED_API_KEY"],
        },
    )
    with urllib.request.urlopen(request, timeout=60) as response:
        return json.load(response)["choices"][0]["message"]["content"]

print(ask_evolved("Name three materials for a sci-fi rocket."))

The answer appears in the system console (Window → Toggle System Console on Windows).

Good to know

The model runs Python inside Blender to make changes, so save your file first; Blender's undo works on what it did. Each step is a request on your Usage page; bigger scenes make bigger requests. Claude models are especially good at 3D work.