Blender Open MCP

Community

Control Blender 3D with natural language via local AI models

How to install ↓ Compare GitHub source repository ↗

Category: Dev Tools #355 of 438
Stars: ★ 120 Popular
Transport: stdio (local process) · Streamable HTTP
Runtime: Python 3.10+ · Prebuilt binary
Credentials: No credential needed
License: MIT
Last commit: 37 days ago
Tools: 16

40FMRS · D

This server provides a practical MCP integration to control Blender via natural language through local Ollama models, suitable for developers and 3D artists. Its architecture is clear and tools are rich, including scene operations, materials, rendering, and AI assistance. However, executing arbitrary code and deleting objects pose security risks, requiring caution. It is non-official, depends on local models, and is not suitable for users needing cloud AI or low security risk.

Strongest · Documentation 13/20 Weakest · Security and permissions 3/20

Reliability

6/20

Security and permissions

3/20

Maintenance

6/20

Documentation

13/20

Setup experience

12/20

Why each score

Reliability 6/20: Evidence is limited to the README: it describes a three-process architecture (FastMCP server, Blender add-on, Ollama), lists 15 tools, CLI/Python clients, and default ports. Deductions: no repository source files, CI workflow, or committed tests were supplied, so the MCP init handshake, tool-list agreement, and real tool behavior cannot be verified; the README itself contains a malformed Ollama command (`ollama pull ollama run llama3.2`) and lacks error-path documentation. Under static calibration, a high score requires verifiable execution evidence, which is absent.
Security and permissions 3/20: Red-line risk is present: `blender_execute_code` runs arbitrary Python/bpy code inside Blender and `blender_delete_object` deletes scene objects, with no confirmation mechanism, least-privilege boundary, or permission scoping documented. PolyHaven downloads also add an external network surface. Therefore security is capped at the 0-4 band and scored 3. No credential handling was present to evaluate, and install examples contained no real tokens or secrets.
Maintenance 6/20: Evidence: the repository is not archived, has 109 stars (used only as a discovery signal) and 13 open issues; the README claims an MIT license. Deductions: metadata shows License NOASSERTION, so no confirmed LICENSE file is evident; there are no releases, commit cadence, dependency update policy, security-response channel, or evidence of issue responsiveness. Sustained maintenance and clear ownership cannot be established from the supplied material.
Documentation 13/20: Evidence: the README provides an architecture diagram, prerequisites, Blender add-on installation, server/client usage, a tool table with modification flags, default ports, development commands, and a troubleshooting table. Deductions: per-tool parameter schemas, error codes, auth/security guidance, and resource/cost limits are missing; documentation is single-layered and not cross-checked against actual behavior, so full marks are not justified.
Setup experience 12/20: Evidence: setup steps are clear and include uv environment creation, Blender add-on installation, Ollama model pull, server launch commands, and mcp.json examples for Claude Desktop/Cursor. Deductions: the setup requires manually running and coordinating three processes (Blender, Ollama, MCP server), the Blender add-on installation depends on GUI actions, and no CI-verified smoke test or platform-specific guidance is provided; static calibration caps setup at 15 without verifiable execution evidence, so 12 is awarded.

Static review · not runListed 2026-08-07

Read the FMRS scoring method →

Fit and risk

What it can accessWrites / deletes local filesRuns commands or codeUses the network

Best for

3D artists who want to control Blender with local AI models
Developers who need to automate 3D workflows
Users integrating with Claude Desktop or Cursor for AI-assisted modeling

Not for

Users without local Ollama or ability to run local LLMs
Users needing cloud AI services (since it uses local models)
Users unfamiliar with Blender add-on installation
Users requiring official support (non-official project)

Required permissions

Can read and modify Blender scene (objects, materials, transforms)
Can execute arbitrary Python/bpy code via blender_execute_code
Can delete objects via blender_delete_object
Can render images to files
Can access internet to download PolyHaven assets
Can access local models via Ollama
Requires network ports (8000, 9876, 11434)

Risks and side effects

Executing arbitrary code may corrupt scenes or systems if given malicious prompts
Deleting objects is irreversible and may lose work
PolyHaven downloads require internet and may fail
Local LLM may generate inaccurate code; review before execution
Non-official project may have bugs or security vulnerabilities

Setup

Before you start

Runtime：Python 3.10+ · Prebuilt binary

Install dependencies: Blender 3.0+, Python 3.10+, Ollama, uv.
Clone repo and set up virtual env: uv venv, activate, uv pip install -e ..
Install Blender add-on: Edit → Preferences → Add-ons → Install, select addon.py, enable "Blender MCP", start MCP server in N-sidebar (default port 9876).
Pull an Ollama model: ollama pull ollama run llama3.2.
Start Ollama server.
Start MCP server: blender-mcp, or with custom options like --transport stdio.

claude_desktop_config.json

{
  "mcpServers": {
    "blender-open-mcp": {
      "command": "blender-mcp",
      "args": ["--transport", "stdio"]
    }
  }
}

Shown for Claude Desktop. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json

{
  "servers": {
    "blender-open-mcp": {
      "command": "blender-mcp",
      "args": [
        "--transport",
        "stdio"
      ]
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal

claude mcp add blender-open-mcp -- blender-mcp --transport stdio

Run it in a terminal; replace any <…> placeholders with your own values first.

Generate and validate client config

Check that it works

Start the add-on's MCP Server inside Blender and run ollama serve, then launch blender-mcp --transport stdio; blender_get_scene_info should appear in the client's tool list, and calling it should return scene info, proving the connection works.

Troubleshooting

Cannot connect to Blender add-on: start MCP server in Blender sidebar
Cannot connect to Ollama: run `ollama serve`
Object not found: check exact name via `blender_get_scene_info`
Render fails: ensure output directory exists and is writable
PolyHaven download fails: check internet connection; try lower resolution

Things to try

Once connected, you can ask your AI assistant things like:

Show me what objects are in the current scene
Create a metallic sphere at position 0, 0, 2
Apply a brick wall texture to MyCube
Render the current scene and save it as an image

Tools 16

blender_get_scene_info read-only

Full scene summary: objects, camera, render settings

blender_get_object_info read-only

Detailed object info: transforms, materials, mesh stats

blender_create_object writes

Add a primitive mesh (CUBE, SPHERE, CYLINDER, ...)

blender_modify_object writes

Change location, rotation, scale, visibility

blender_delete_object destructive

Remove an object from the scene

blender_set_material writes

Create and assign a Principled BSDF material

blender_render_image writes

Render current scene to a file

blender_execute_code destructive

Run arbitrary Python/bpy code in Blender

Show 8 more tools

blender_get_polyhaven_categories read-only

List PolyHaven asset categories

blender_search_polyhaven_assets read-only

Search PolyHaven library with pagination

blender_download_polyhaven_asset writes

Download & import a PolyHaven asset

blender_set_texture writes

Apply a downloaded PolyHaven texture to an object

blender_ai_prompt read-only

Send a natural language prompt to Ollama

blender_get_ollama_models read-only

List available local Ollama models

blender_set_ollama_model writes

Switch the active Ollama model

blender_set_ollama_url writes

Update the Ollama server URL

Use cases

Create and modify 3D objects in Blender via natural language commands

Use AI assistant to generate and execute bpy code

Download and apply textures from PolyHaven

Automate scene rendering

Get scene and object information for AI analysis

Supported clients

Claude Desktop

Cursor

Listed from the project's documentation, not tested by this site.

Overview

Built on the Model Context Protocol (MCP), connecting Claude, Cursor, or any MCP client to Blender through a local Ollama LLM. Architecture uses FastMCP server, Blender add-on, and Ollama.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 62.9k · Tools 2 Compare with this →

NPM Sentinel MCP Server 79 · B

AI-powered NPM package analysis MCP server

★ 18 · Tools 19 Compare with this →

MartinLoop MCP Server 79 · B

An independent control layer that gives AI coding agents budgets, brakes, and receipts

★ 410 Compare with this →

GitHub MCP Server 76 · B

GitHub's official MCP server for managing repos, issues, PRs, and workflows via natural language

★ 33.5k · Tools 15 Compare with this →

Compare with other MCP servers

Source revision a98c5a8e9cec Data synced 2026-10-12 Read the FMRS scoring method