Godot MCP (Model Context Protocol)

Project Overview

Godot MCP provides a comprehensive, production-ready integration between Large Language Models (LLMs) and the Godot Engine (versions 4.x). It creates a unified Model Context Protocol interface allowing AI assistants to intelligently read, write, and manipulate Godot project files, interact with the live Godot Editor, and perform End-to-End (E2E) testing on running games. By consolidating multiple AI-to-Godot features into a single robust toolkit, Godot MCP dramatically lowers the friction of AI-assisted game development.

Installation & Usage

1. Prerequisites

• Node.js (v18 or higher)
• Godot Engine 4.x
• Supported MCP client (e.g., Claude Desktop)

2. Add the Godot Plugin

Navigate to your specific Godot project's root folder in your terminal and run:

npx -y @xianlee/godot-mcp install

This will automatically copy the addons/godot_mcp folder into your project. Enable the godot_mcp plugin via Project > Project Settings > Plugins in the Godot Editor.

3. Configure the MCP Client (Claude Desktop Example)

Add the following configuration to your mcp config json:

{
  "mcpServers": {
    "godot-mcp": {
      "command": "npx",
      "args": ["-y", "@xianlee/godot-mcp"],
      "env": {
        "GODOT_PATH": "<OPTIONAL_ABSOLUTE_PATH_TO_GODOT_EXECUTABLE>"
      }
    }
  }
}

4. Running the Complete Toolchain

1. Ensure your Godot editor is open with the godot_mcp plugin enabled.
2. Start your MCP Client (like Claude Desktop). The server will download on-the-fly and seamlessly jump into the Godot WebSocket.

Feature List

• Live Editor Manipulation: Modify nodes, properties, and scenes in real-time within the Godot Editor.
• Deep Project Context: Read and analyze scripts, scene structures (.tscn), and project files dynamically.
• Version-Aware Godot Docs: Automatically query Godot documentation matching your specific engine version.
• End-to-End Game Bridge testing: Play, test, and assert logic in live running games via simulated inputs and state queries.
• Advanced Diagnostics: View diagnostics, validate scenes, and debug live errors automatically.
• Consolidated Architecture: Minimal tool footprint heavily utilizing action parameters to conserve AI context windows.
• Auto-Screenshot Loop: Periodically captures visual screenshots of the editor or running game state to provide AI with continuous visual context.
• Bug Fix Loop: Automatically collects stack traces from the Godot engine, analyzes runtime errors, and iteratively proposes code fixes.
• Scene Tree Viewer: Interactive web-based visualizer for inspecting the live Godot scene tree and node properties via a browser.
• Intelligent Path Detection: Seamlessly detects correct Godot project paths and environment variables, even when loaded via external IDEs.
• Robust Connection Handling: Enterprise-grade WebSocket connections featuring command queuing, exponential backoff, and large payload buffering.
• Universal Asset Resolution: Resolves internal Godot UUIDs (uid://... or .uid files) dynamically to absolute file paths, ensuring seamless AI context mapping.

MCP Tools Reference

Acknowledgements

This project builds upon the foundational work of several existing Model Context Protocol (MCP) implementations for the Godot Engine. We would like to express our gratitude to the original maintainers and communities behind these repositories:

• godot-mcp-plugin: (tomyud1/godot-mcp)
• godot-mcp-scene: (Derfirm/godot-mcp)
• godot-mcp-docs: (Nihilantropy/godot-mcp-docs)
• godot-mcp-solo: (Coding-Solo/godot-mcp)
• godot_mcp_uid: (bradypp/godot-mcp)

We sincerely thank the original maintainers, contributors, and the broader Godot and open-source communities who laid the groundwork for these tools. Their specialized prototypes proved the viability of connecting Large Language Models to Godot via MCP and provided essential insights that made this unified, production-ready toolchain possible.