A Model Context Protocol (MCP) server for interacting with the Godot game engine. It lets AI agents (Claude, Cursor, Cline, Codex, opencode, etc.) inspect, modify, run, and debug a Godot project directly.
This project is a streamlined and extended fork of the popular . It keeps the shared architecture and adds visual verification, project discovery, and scene/script analysis tools on top.Thank you for making this possible with your excellent open-source work!
A headless MCP server for Godot Engine, providing tools to launch the editor, run projects, debug errors, capture screenshots and runtime output, design and manipulate scenes, and automate development workflows through AI coding agents.
Windows note: Always provide a full absolute path, e.g. "GODOT_PATH": "C:\\Godot\\Godot_v4.7.1\\Godot_v4.7.1-stable_win64.exe". Auto-detection scans common locations if godot is not on PATH, but setting the variable yourself is the reliable approach.
Tools
Editor & Process Management
Tool
Description
Inputs
launch_editor
Opens the Godot editor for a specific project.
projectPath
quit_godot
Closes the editor that this server launched via launch_editor.
—
view_log
Returns the console output of the last launched editor.
lineCount (optional, default 50, max 1000)
get_godot_version
Returns the installed Godot version.
—
Running & Debugging
Tool
Description
Inputs
run_project
Runs the project in debug mode and captures output.
projectPath, scene (optional)
run_scene
Runs a single scene (F6-style) with an auto-stop timeout.
Validation tool inputs use camelCase names as shown. The equivalent snake_case aliases project_path, script_path, script_paths, timeout_ms, max_files, and max_output_bytes are also accepted and normalized automatically. maxFiles is capped at 1000, and maxOutputBytes is a combined stdout/stderr cap capped at 256 KiB.
Scene Editing
Tool
Description
Inputs
create_scene
Creates a new scene with the given root node type.
Visual tools require a real display (they are not --headless). On headless Linux, wrap Godot with xvfb-run (see below).
UID Management (Godot 4.4+)
Tool
Description
Inputs
get_uid
Returns the UID for a specific file.
projectPath, filePath
update_project_uids
Updates UID references by resaving resources.
projectPath
How It Works / Architecture
The server uses two primary approaches:
Direct commands — simple operations such as launch_editor, run_project, and get_godot_version use Godot's built-in CLI arguments directly (--editor, --path, --version).
Bundled GDScript operations — complex operations such as create_scene, add_node, load_sprite, and get_scene_insights are executed by two GDScript files that are copied into build/scripts/ during build:
godot_operations.gd — packs/saves scenes, adds nodes, and writes resources.
godot_insights.gd — parses .tscn and .gd files as plain text (no ResourceLoader needed, so it also works in headless environments).
Both scripts receive an operation name and parameters as JSON and print the result to stdout, avoiding temporary files per operation.
The capture_screenshot tools run Godot without--headless to produce real pixels and return the PNG as a base64 image block.
Security Notes
Path validation blocks .. traversal and NUL byte (poison-null-byte) attacks.
A headless MCP server for Godot Engine, providing tools to launch the editor, run projects, debug errors, capture screenshots and runtime output, design and manipulate scenes, and automate development
A headless MCP server for Godot Engine, providing tools to launch the editor, run projects, debug errors, capture screenshots and runtime output, design and manipulate scenes, and automate development workflows through AI coding agents.
도구
도구
아직 도구 목록이 없습니다. 원격 서버는 첫 상태 확인 후에 도구 목록이 생깁니다.
FAQ
질문
Godot MCP Flash은(는) 어떻게 설치하나요?
이 페이지에서 Cursor, Claude Desktop, Claude Code 또는 VS Code용 stdio 설정을 복사합니다. PlugMCP 프록시가 아니라 게시자의 설치 방식을 사용합니다.
마지막 확인: 2026-10-28. A headless MCP server for Godot Engine, providing tools to launch the editor, run projects, debug errors, capture screenshots and runtime output, design and manipulate scenes, and automate development workflows through AI coding agents.