Godot MCP
About
A Model Context Protocol (MCP) server for interacting with the Godot game engine.
Details
- License
- MIT license
Explore
- Launch Godot editor and run projects with debug output capture
- Create, modify, and manage Godot scenes and nodes
- Load textures into Sprite2D nodes and export MeshLibrary resources
- Get Godot version, project info, and list projects in directories
- Support for Godot 4.4+ UID management (get and update UIDs)
- Optional read-only mode for safe project analysis
- Zero-config automatic Godot detection with manual override
npm install
Add to your Cline MCP settings file:
``json
{
"mcpServers": {
"godot": {
"command": "node",
"args": ["/absolute/path/to/godot-mcp/build/index.js"],
"env": {
"DEBUG": "false",
"READ_ONLY": "false",
"GODOT_PATH": "/path/to/godot"
},
"disabled": false,
"autoApprove": [
"launch_editor",
"run_project",
"get_debug_output",
"stop_project",
"get_godot_version",
"list_projects",
"get_project_info",
"create_scene",
"add_node",
"edit_node",
"remove_node",
"load_sprite",
"export_mesh_library",
"save_scene",
"get_uid",
"update_project_uids"
]
}
}
}
godot
1. Open Cursor Settings → Features → MCP
2. Click + Add New MCP Server
3. Configure:
- Name: command
- Type: node /absolute/path/to/godot-mcp/build/index.js
- Command:
4. Click Add and refresh the server list
Create .cursor/mcp.json in your project root:json
{
"mcpServers": {
"godot": {
"command": "node",
"args": ["/absolute/path/to/godot-mcp/build/index.js"],
"env": {
"DEBUG": "false",
"GODOT_PATH": "/path/to/godot",
"READ_ONLY_MODE": "false"
}
}
}
}
bashGODOT_PATH
| Variable | Description | Default | Example |
| ---------------- | -------------------------------- | ------------- | ----------------- |
|| Path to Godot executable | Auto-detected |/usr/bin/godot4|DEBUG
|| Enable detailed logging |false|true|READ_ONLY_MODE
|| Restrict to read-only operations |false|true|DEBUG=true`If you encounter issues not covered here:
1. Check debug logs with
2. Search existing issues on GitHub
3. Create a detailed issue report with:
- Operating system and version
- Node.js and Godot versions
- Complete error messages
- Configuration used
- Steps to reproduce
Tools are registered in the ToolRegistry with metadata indicating their capabilities:
``typescript``
export interface ToolRegistration {
definition: ToolDefinition;
handler: (args: any) => Promise<ToolResponse>;
readOnly: boolean;
}
The registry automatically filters tools based on the current mode (read-only vs. full access) and provides a unified interface for tool discovery and execution.
Table of Contents
- What is Godot MCP? - Features - Requirements - Installation - Configuration - API Reference - Project Architecture - Usage Examples - Read-Only Mode - Troubleshooting - FAQ - Contributing - LicenseWhat is Godot MCP?
Godot MCP bridges the gap between AI assistants and the Godot game engine by providing a standardized Model Context Protocol interface. This powerful integration enables AI assistants like Claude, Cursor, and Cline to directly interact with Godot projects through a comprehensive set of tools.Key Value Propositions
- Direct Godot Integration: Launch editors, run projects, and capture debug output programmatically - Scene Management: Create, modify, and manage Godot scenes through AI commands - Real-time Feedback: AI assistants can see actual Godot output and errors for better assistance - Cross-platform Compatibility: Works seamlessly on Windows, macOS, and Linux - Secure Operations: Optional read-only mode for safe project analysis - Zero Configuration: Automatic Godot detection with manual override optionsHow It Works
The server acts as a middleware layer between your AI assistant and Godot, translating natural language commands into specific Godot operations. When you ask your AI to "create a player scene with a sprite," the MCP server: 1. Validates the request and project structure 2. Executes the appropriate Godot operations 3. Returns detailed success/error feedback 4. Enables the AI to understand and respond to the results This creates a powerful feedback loop where AI assistants can learn from actual Godot behavior, leading to more accurate code generation and debugging assistance.Features
Core Project Management
- 🚀 Launch Godot Editor: Open the Godot editor for specific projects - ▶️ Run Godot Projects: Execute projects in debug mode with real-time output capture - 🛑 Control Execution: Start and stop Godot projects programmatically - 📊 Debug Output Capture: Retrieve comprehensive console output and error messages - ℹ️ System Information: Get installed Godot version and project metadata - 📁 Project Discovery: Find and list Godot projects in specified directoriesAdvanced Scene Management
- 🎬 Create New Scenes: Generate scenes with specified root node types - ➕ Add Nodes: Insert nodes into existing scenes with customizable properties - ✏️ Edit Node Properties: Modify positions, scales, textures, and other node attributes - 🗑️ Remove Nodes: Clean up scenes by removing unwanted nodes - 🖼️ Load Sprites: Automatically load textures into Sprite2D nodes - 🧱 Export MeshLibrary: Convert 3D scenes to MeshLibrary resources for GridMap - 💾 Save Scene Variants: Create scene copies and manage scene versionsGodot 4.4+ UID Management
- 🔗 Get File UIDs: Retrieve unique identifiers for project resources - 🔄 Update UID References: Maintain proper resource links during project upgradesSecurity & Safety
- 🔒 Read-Only Mode: Restrict operations to analysis-only for secure environments - ✅ Path Validation: Comprehensive project and file path verification - 🛡️ Error Handling: Robust error reporting with actionable suggestionsRequirements
System Requirements
- Godot Engine: Version 3.5+ or 4.0+ (latest stable recommended) - Node.js & npmAI Assistant Compatibility
- Cline & Roo Code: Full support with auto-approval configuration - Cursor & VS Code: Supports both UI and project-specific configuration - Claude Desktop: Compatible with MCP server integration - Other MCP-enabled tools: Any tool supporting the Model Context ProtocolInstallation
Clone and Build
```bashSign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.



