mcp-srver-starter-pack by namurokuro - MCP Server

Blender Evolving MCP generator

Model Context Protocol (MCP) server for integrating Blender-Ollama system with Cursor.

📚 Repository Index: See for complete documentation index and navigation guide.

Overview

This MCP server exposes the Blender-Ollama specialized agent system to Cursor, allowing you to:

Keep a constantly evolving workspace that individuals can tailor to their own pipelines—use the Starter Pack deliverables as stable references while experimenting freely with learning workflows, AI generators, or new agent behaviors.
Create 3D scenes in Blender via natural language
Query operation history from 11 specialized databases
Access code patterns and performance metrics
Use 14 specialized agents for different Blender domains
Monitor all agent activities in real-time via web viewport (see )

Starter Pack (AI-ready)

Need a lightweight bundle for onboarding, demos, or source sharing? Regenerate the curated Starter Pack anytime:

python build_starter_pack.py
Use the starter_pack/ folder as the upload artifact (Git, ZIP, etc.).

The Starter Pack keeps:

Final/future-proof documentation (all FINAL_*, COMPLETE_*, READY* files)
AI generator entrypoints such as finalize_vape_ad.py and render_final.py
Ready-to-run configs (cursor_mcp_config_ready.json, Docker-ready equivalent)

Together those files represent the full functionality showcase—users can study the docs, run the generator scripts, or extend them with their own models without needing the rest of the experimental workspace.

Architecture

Cursor → MCP Server (stdio) → Agent Coordinator → Specialized Agents → Blender
                                              ↓
                                        11 SQLite Databases
                                         ↓
                              Agent Activity Tracker → Viewport Server (Web)

Agent Activity Viewport

Monitor all agent activities in real-time through a web-based dashboard:

Start viewport: start_viewport.bat (Windows) or ./start_viewport.sh (Linux/Mac)
Access dashboard: http://localhost:5000
See real-time agent status, operations, progress, and activity logs
Full documentation:

Installation

Option 1: Direct Installation (Recommended for Development)

Ensure the parent blender-ollama directory is accessible
The server automatically imports from the parent directory
Install dependencies: pip install -r requirements.txt

Option 2: Docker Installation (Recommended for Production)

Prerequisites: Docker and Docker Compose installed
Migrate from direct installation: migrate-to-docker.bat (Windows)
Start services: docker-start.bat (Windows) or ./docker-start.sh (Linux/Mac)
Pull Ollama models: docker-pull-models.bat
Check status: docker-status.bat

For detailed Docker setup, see:

- Migration guide (start here!)
- Quick start guide
- Complete documentation

Configuration

Cursor Configuration

Add to your Cursor settings (.cursor/mcp.json or Cursor settings):

{
  "mcpServers": {
    "blender-ollama": {
      "command": "python",
      "args": [
        "F:/mcp server/mcp_server.py"
      ],
      "env": {
        "OLLAMA_URL": "http://localhost:11434",
        "BLENDER_HOST": "localhost",
        "BLENDER_PORT": "9876"
      }
    }
  }
}

Available Tools

1. `create_scene`

Create a 3D scene in Blender from natural language description.

Parameters:

description (required): Natural language description
field (optional): Specialist agent to use

Example:

{
  "name": "create_scene",
  "arguments": {
    "description": "Create a red cube on a blue plane",
    "field": "modeling"
  }
}

2. `get_scene_info`

Get current Blender scene information.

3. `execute_blender_code`

Execute Python code directly in Blender.

4. `query_database`

Query operation history, patterns, errors, or performance.

Parameters:

database: Which database to query (or "all")
query_type: "recent", "patterns", "errors", or "performance"
limit: Maximum results

5. `get_model_performance`

Get LLM model performance metrics.

6. `get_successful_patterns`

Get successful code generation patterns.

7. `list_specialists`

List all available specialist agents.

8. `get_development_proposals`

Get development proposals based on current trends and innovations. Monitors trends in Blender, AI, video editing, fashion, furniture, TikTok, Instagram, gaming, and other project-relevant areas. Automatically adapts to your current project context.

Parameters:

focus_area (optional): "general", "blender", "ai", "tech", "video", "fashion", "furniture", "tiktok", "instagram", "gaming", or "custom"
custom_topics (optional): Array of custom topics for project-specific analysis
use_project_context (optional): Use current project context to adapt proposals (default: true)

9. `set_project_context`

Set your current project context so trend monitoring adapts to your specific project type.

Parameters:

project_type (required): "fashion", "furniture", "video", "tiktok", "instagram", "gaming", "blender", "3d", "modeling", or "custom"
project_description (optional): Description of your project

10. `get_project_relevant_trends`

Get trends automatically adapted to your current project context. No parameters needed - uses your set project context.

Available Resources

Resources provide read-only access to data:

blender://database/{field}/schema - Database schema
blender://database/{field}/operations - Recent operations
blender://database/{field}/patterns - Code patterns
blender://database/{field}/errors - Error patterns
blender://database/{field}/performance - Performance metrics
blender://scene/current - Current Blender scene
blender://agents/list - Available agents

Available Prompts

create_modeling_scene - Create modeling scene workflow
create_material_setup - Material setup workflow
analyze_performance - Performance analysis workflow
find_similar_operations - Find similar operations

Testing

Using MCP Inspector

npx @modelcontextprotocol/inspector python "F:/mcp server/mcp_server.py"

Manual Testing

python "F:/mcp server/mcp_server.py"

Then send JSON-RPC requests via stdin.

Specialized Agents

The server routes tasks to 10 specialized agents:

Modeling - 3D modeling and mesh operations
Shading - Materials and shaders (includes Sanctus Library procedural shaders support)
Animation - Animation and keyframes
VFX - Visual effects and simulations
Motion Graphics - Text and motion graphics
Rendering - Rendering and export
Rigging - Armatures and rigging
Sculpting - Digital sculpting
Camera Operator - Camera operations
Videography - Video editing

Databases

Each agent maintains its own SQLite database:

modeling_data.db
shading_data.db
animation_data.db
vfx_data.db
motiongraphics_data.db
rendering_data.db
rigging_data.db
sculpting_data.db
cameraoperator_data.db
videography_data.db

Troubleshooting

Import Errors

Ensure the parent blender-ollama directory is accessible
Check that specialized_agents.py and data_collector.py exist
Verify Python path includes the parent directory

Connection Errors

Ensure Ollama is running on localhost:11434
Ensure Blender addon is running on localhost:9876
Check firewall settings

Database Errors

Ensure database files exist in the parent directory
Check file permissions
Verify database paths are correct

Sanctus Library Integration

The Shading agent now supports Sanctus Library procedural shaders collection, providing access to 690+ high-quality procedural materials.

Installation

Purchase and download Sanctus Library from: https://superhivemarket.com/products/sanctus-library-addon---procedural-shaders-collection-for-blender/
Install in Blender:
- Edit > Preferences > Add-ons
- Click "Install..." and select the Sanctus Library .zip file
- Enable the addon
Access materials through Asset Browser (Shift+A) or use the Python API

Usage

Via MCP Server:

{
  "name": "create_scene",
  "arguments": {
    "description": "Apply Sanctus Library metal material to cube",
    "field": "shading"
  }
}

Via Python Script:

from sanctus_library_tools import apply_sanctus_material_to_object

# Apply material to object
result = apply_sanctus_material_to_object("Cube", "MetalMaterial")

Example Scripts:

use_sanctus_library.py - Check installation and list materials
example_sanctus_materials.py - Create scene with Sanctus materials

Available Functions

The sanctus_library_tools.py module provides:

check_sanctus_installed() - Check if addon is installed
apply_sanctus_material_to_object() - Apply material to object
get_sanctus_materials() - List available materials
get_sanctus_material_categories() - Get material categories
Code generation functions for programmatic material application

References

Protocol & Integration

Cursor MCP Server Guide - Building MCP servers for Cursor
MCP Protocol Specification - Model Context Protocol specification

Blender Documentation

Blender Python API - Official Blender Python API reference
Blender Developer Documentation - Blender development handbook and architecture
Blender Features Documentation - Design and implementation of Blender features
Blender Projects - Blender source code, issues, and development platform
Blender Developer Forum - Community forum for Blender developers
Blender Release Notes - API changes and compatibility notes

License

Same as parent Blender-Ollama project.

namurokuro/mcp-srver-starter-pack

Blender Evolving MCP generator

Overview

Starter Pack (AI-ready)

Architecture

Agent Activity Viewport

Installation

Option 1: Direct Installation (Recommended for Development)

Option 2: Docker Installation (Recommended for Production)

Configuration

Cursor Configuration

Available Tools

1. create_scene

2. get_scene_info

3. execute_blender_code

4. query_database

5. get_model_performance

6. get_successful_patterns

7. list_specialists

8. get_development_proposals

9. set_project_context

10. get_project_relevant_trends

Available Resources

Available Prompts

Testing

Using MCP Inspector

Manual Testing

Specialized Agents

Databases

Troubleshooting

Import Errors

Connection Errors

Database Errors

Sanctus Library Integration

Installation

Usage

Available Functions

References

Protocol & Integration

Blender Documentation

License

1. `create_scene`

2. `get_scene_info`

3. `execute_blender_code`

4. `query_database`

5. `get_model_performance`

6. `get_successful_patterns`

7. `list_specialists`

8. `get_development_proposals`

9. `set_project_context`

10. `get_project_relevant_trends`