Control Godot Engine with AI
MCP server letting AI agents launch, run, debug, and manipulate Godot game projects - scenes, nodes, execution control.
Maintainer of this project? Claim this page to edit the listing.
1.0.0Add to Favorites
Why it matters
Enable AI assistants to interact with the Godot game engine, automating tasks like launching the editor, executing projects, and managing scenes.
Outcomes
What it gets done
Launch and manage Godot editor instances.
Execute and debug Godot projects programmatically.
Create, modify, and save game scenes.
Retrieve project information and manage UIDs.
Install
Add it to your toolbox
Run in your project directory:
curl -fsSL https://spark.entire.vc/get/vb-godot | bash Capabilities
Tools your agent gets
Launch the Godot editor for a specific project
Execute Godot projects in debug mode
Retrieve console output and error messages
Stop running Godot projects
Get the version of installed Godot
Find Godot projects in a specified directory
Get detailed information about project structure
Create new scenes with specified root node types
Overview
Godot MCP Server
An MCP server letting AI agents launch, run, and debug Godot game projects, with tools for scene/node management, sprite loading, and UID handling. Use when an AI agent needs to interact with an actual Godot project - launching, running, debugging, or editing scenes.
What it does
Godot MCP lets AI agents launch the Godot editor, run projects in debug mode, capture console/debug output, and control execution, giving a direct feedback loop for understanding what works in a real Godot project and enabling better code generation and debugging assistance. It also supports scene management (create scenes, add nodes, load sprites/textures, export MeshLibrary resources, save scene variants) and UID management for Godot 4.4+.
When to use - and when NOT to
Use it when an AI agent needs to interact with an actual Godot project - launching the editor, running/debugging a scene, inspecting project structure, or manipulating scenes and nodes directly - rather than only generating GDScript as static text. Requires the Godot Engine installed locally, Node.js >=18.0.0 with npm, and an MCP-capable AI agent.
Capabilities
Named tools/features: launch the Godot editor for a project, run a project in debug mode, capture debug output (console/errors), start/stop execution, get the installed Godot version, list Godot projects in a directory, get detailed project structure info, create new scenes with a specified root node type, add nodes to existing scenes with custom properties, load sprites/textures into Sprite2D nodes, export 3D scenes as MeshLibrary resources for GridMap, save scenes (with variant options), and (Godot 4.4+) get a file's UID or update UID references by resaving resources. Internally, simple operations (launching the editor, getting project info) call Godot's built-in CLI directly, while complex operations (creating scenes, adding nodes) run through a single bundled GDScript file (godot_operations.gd) that accepts operation type and parameters as JSON, avoiding temporary per-operation files.
How to install
For Claude Code:
claude mcp add godot -- npx @coding-solo/godot-mcp
Environment variables can be passed the same way (-e GODOT_PATH=/path/to/godot -e DEBUG=true). Cline, Cursor, and other MCP clients register it via an mcpServers entry running npx @coding-solo/godot-mcp, with GODOT_PATH (overrides automatic Godot executable detection) and DEBUG ("true" for detailed server-side logging) as the two supported environment variables. To build from source instead of using npx: git clone, npm install, npm run build, then point the client at build/index.js. Named troubleshooting fixes: set GODOT_PATH if Godot isn't found, restart the AI assistant for connection issues, verify the project path contains a project.godot file, and run npm install for build issues; in Cursor specifically, MCP tools only run under the Agent chat profile (Pro/Business subscription), and "Yolo Mode" auto-runs MCP tool requests. Licensed under the MIT License.
Who it's for
Godot developers who want an AI agent to actually run, debug, and manipulate their game project - not just write GDScript blind - closing the feedback loop between code generation and real editor/runtime behavior.
Source README
Godot MCP
((((((( (((((((
((((((((((( (((((((((((
((((((((((((( (((((((((((((
(((((((((((((((((((((((((((((((((
(((((((((((((((((((((((((((((((((
((((( ((((((((((((((((((((((((((((((((((((((((( (((((
(((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((
((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((
((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((
(((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((
(((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((
(((((((((((@@@@@@@(((((((((((((((((((((((((((@@@@@@@(((((((((((
(((((((((@@@@,,,,,@@@(((((((((((((((((((((@@@,,,,,@@@@(((((((((
((((((((@@@,,,,,,,,,@@(((((((@@@@@(((((((@@,,,,,,,,,@@@((((((((
((((((((@@@,,,,,,,,,@@(((((((@@@@@(((((((@@,,,,,,,,,@@@((((((((
(((((((((@@@,,,,,,,@@((((((((@@@@@((((((((@@,,,,,,,@@@(((((((((
((((((((((((@@@@@@(((((((((((@@@@@(((((((((((@@@@@@((((((((((((
(((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((
(((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((
@@@@@@@@@@@@@((((((((((((@@@@@@@@@@@@@((((((((((((@@@@@@@@@@@@@
((((((((( @@@(((((((((((@@(((((((((((@@(((((((((((@@@ (((((((((
(((((((((( @@((((((((((@@@(((((((((((@@@((((((((((@@ ((((((((((
(((((((((((@@@@@@@@@@@@@@(((((((((((@@@@@@@@@@@@@@(((((((((((
(((((((((((((((((((((((((((((((((((((((((((((((((((((((((((
(((((((((((((((((((((((((((((((((((((((((((((((((((((
(((((((((((((((((((((((((((((((((((((((((((((((
(((((((((((((((((((((((((((((((((
/$$ /$$ /$$$$$$ /$$$$$$$
| $$$ /$$$ /$$__ $$| $$__ $$
| $$$$ /$$$$| $$ \__/| $$ \ $$
| $$ $$/$$ $$| $$ | $$$$$$$/
| $$ $$$| $$| $$ | $$____/
| $$\ $ | $$| $$ $$| $$
| $$ \/ | $$| $$$$$$/| $$
|__/ |__/ \______/ |__/
A Model Context Protocol (MCP) server for interacting with the Godot game engine.
Introduction
Godot MCP enables AI agents to launch the Godot editor, run projects, capture debug output, and control project execution. This direct feedback loop helps agents understand what works and what doesn't in real Godot projects, leading to better code generation and debugging assistance.
Features
- Launch Godot Editor: Open the Godot editor for a specific project
- Run Godot Projects: Execute Godot projects in debug mode
- Capture Debug Output: Retrieve console output and error messages
- Control Execution: Start and stop Godot projects programmatically
- Get Godot Version: Retrieve the installed Godot version
- List Godot Projects: Find Godot projects in a specified directory
- Project Analysis: Get detailed information about project structure
- Scene Management:
- Create new scenes with specified root node types
- Add nodes to existing scenes with customizable properties
- Load sprites and textures into Sprite2D nodes
- Export 3D scenes as MeshLibrary resources for GridMap
- Save scenes with options for creating variants
- UID Management (for Godot 4.4+):
- Get UID for specific files
- Update UID references by resaving resources
Requirements
- Godot Engine installed on your system
- Node.js (>=18.0.0) and npm
- An AI agent that supports MCP
Quick Start
Claude Code
claude mcp add godot -- npx @coding-solo/godot-mcp
That's it. Restart Claude Code and your Godot MCP tools are available.
With environment variables:
claude mcp add godot -e GODOT_PATH=/path/to/godot -e DEBUG=true -- npx @coding-solo/godot-mcp
Cline
Add to your Cline MCP settings file (~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json):
{
"mcpServers": {
"godot": {
"command": "npx",
"args": ["@coding-solo/godot-mcp"],
"env": {
"DEBUG": "true"
},
"disabled": false,
"autoApprove": [
"launch_editor",
"run_project",
"get_debug_output",
"stop_project",
"get_godot_version",
"list_projects",
"get_project_info",
"create_scene",
"add_node",
"load_sprite",
"export_mesh_library",
"save_scene",
"get_uid",
"update_project_uids"
]
}
}
}
Cursor
Using the Cursor UI:
- Go to Cursor Settings > Features > MCP
- Click on the + Add New MCP Server button
- Fill out the form:
- Name:
godot - Type:
command - Command:
npx @coding-solo/godot-mcp
- Name:
- Click "Add"
- You may need to press the refresh button in the top right corner of the MCP server card to populate the tool list
Using Project-Specific Configuration:
Create a file at .cursor/mcp.json in your project directory:
{
"mcpServers": {
"godot": {
"command": "npx",
"args": ["@coding-solo/godot-mcp"],
"env": {
"DEBUG": "true"
}
}
}
}
Other MCP Clients
For any MCP-compatible client, use this configuration:
{
"mcpServers": {
"godot": {
"command": "npx",
"args": ["@coding-solo/godot-mcp"],
"env": {
"GODOT_PATH": "/path/to/godot",
"DEBUG": "true"
}
}
}
}
Environment Variables
| Variable | Description |
|---|---|
GODOT_PATH |
Path to the Godot executable (overrides automatic detection) |
DEBUG |
Set to "true" to enable detailed server-side debug logging |
Building from Source
git clone https://github.com/Coding-Solo/godot-mcp.git
cd godot-mcp
npm install
npm run build
Then point your MCP client to build/index.js instead of using npx.
Architecture
The Godot MCP server uses a bundled GDScript approach for complex operations:
- Direct Commands: Simple operations like launching the editor or getting project info use Godot's built-in CLI commands directly.
- Bundled Operations Script: Complex operations like creating scenes or adding nodes use a single, comprehensive GDScript file (
godot_operations.gd) that handles all operations.
The bundled script accepts operation type and parameters as JSON, allowing for flexible and dynamic operation execution without generating temporary files for each operation.
Troubleshooting
- Godot Not Found: Set the
GODOT_PATHenvironment variable to your Godot executable path - Connection Issues: Ensure the server is running and restart your AI assistant
- Invalid Project Path: Ensure the path points to a directory containing a
project.godotfile - Build Issues: Make sure all dependencies are installed by running
npm install
Cursor-Specific Issues
- Ensure the MCP server shows up and is enabled in Cursor settings (Settings > MCP)
- MCP tools can only be run using the Agent chat profile (Cursor Pro or Business subscription)
- Use "Yolo Mode" to automatically run MCP tool requests
FAQ
Common questions
Discussion
Questions & comments · 0
Sign In Sign in to leave a comment.