Data & analysis

Openclaw Godot Skill

Try it

Drive a Godot 4.x Editor from an agent through 30 built-in tools covering scenes, nodes, transforms, input, screenshots, and logs.

What it does

Sends commands to a running Godot 4.x Editor via a plugin's `godot_execute` interface. Covers scene creation, node creation and deletion across 80+ node types, transforms, property reads and writes with Vector2/Vector3 auto-conversion, input simulation during Play mode, viewport screenshots, scene tree inspection, and console log retrieval. Supports an OpenClaw Gateway connection over WebSocket for remote channels and a local MCP HTTP bridge on port 27183 for local AI tools. Returns editor state, project metadata, screenshots, and log lines. Destructive operations require user confirmation.

When to use it

  • Creating scenes and adding nodes programmatically
  • Driving gameplay tests via Play-mode input simulation
  • Inspecting scene trees and capturing viewport screenshots
  • Reading errors and warnings from Godot log files

The skill document

Godot Plugin Skill

Control Godot 4.x Editor through 30 built-in tools. Supports 80+ node types.

Connection Modes

Mode A: OpenClaw Gateway (Remote)

For Telegram, Discord, and other OpenClaw channels:

  • Plugin connects TO Gateway via WebSocket
  • Gateway URL: configurable in Godot dock UI
  • Best for: remote access, mobile messaging

Mode B: MCP Bridge (Local)

For Claude Code, Cursor, and local AI tools:

  • Built-in HTTP server in Godot Plugin
  • Start: Click "Start MCP" in dock UI (port 27183)
  • Add to Claude Code: claude mcp add godot -- node /MCP~/index.js

First-Time Setup

If godot_execute tool is not available, install the gateway extension:

# From skill directory
./scripts/install-extension.sh

# Restart gateway
openclaw gateway restart

The extension files are in extension/ directory.

Quick Reference

Core Tools

CategoryKey Tools
Scenescene.create, scene.getCurrent, scene.open, scene.save
Nodenode.find, node.create, node.delete, node.getData
Transformtransform.setPosition, transform.setRotation, transform.setScale
Debugdebug.tree, debug.screenshot, console.getLogs
Inputinput.keyPress, input.mouseClick, input.actionPress
Editoreditor.play, editor.stop, editor.getState

Common Workflows

1. Scene Creation

Use godot_execute tool:

  • godot_execute(tool="scene.create", parameters={rootType: "Node2D", name: "Level1"})
  • godot_execute(tool="node.create", parameters={type: "CharacterBody2D", name: "Player"})
  • godot_execute(tool="scene.save")

2. Find & Modify Nodes

  • godot_execute(tool="node.find", parameters={name: "Player"})
  • godot_execute(tool="node.getData", parameters={path: "Player"})
  • godot_execute(tool="transform.setPosition", parameters={path: "Player", x: 100, y: 200})

3. Game Testing with Input

  • godot_execute(tool="editor.play")
  • godot_execute(tool="input.keyPress", parameters={key: "W"})
  • godot_execute(tool="input.actionPress", parameters={action: "jump"})
  • godot_execute(tool="debug.screenshot")
  • godot_execute(tool="editor.stop")

4. Check Logs

  • godot_execute(tool="console.getLogs", parameters={limit: 50})
  • godot_execute(tool="console.getLogs", parameters={type: "error", limit: 20})

Tool Categories

Console (2 tools)

  • console.getLogs - Get logs from Godot log file {limit: 100, type: "error"|"warning"|""}
  • console.clear - Placeholder (logs can't be cleared programmatically)

Scene (5 tools)

  • scene.getCurrent - Get current scene info
  • scene.list - List all .tscn/.scn files
  • scene.open - Open scene by path
  • scene.save - Save current scene
  • scene.create - Create new scene {rootType: "Node2D"|"Node3D"|"Control", name: "SceneName"}

Node (6 tools)

  • node.find - Find by name, type, or group
  • node.create - Create node (80+ types: CSGBox3D, MeshInstance3D, ColorRect, etc.)
  • node.delete - Delete node by path
  • node.getData - Get node info, children, transform
  • node.getProperty - Get property value
  • node.setProperty - Set property value (Vector2/3 auto-converted)

Transform (3 tools)

  • transform.setPosition - Set position {x, y} or {x, y, z}
  • transform.setRotation - Set rotation (degrees)
  • transform.setScale - Set scale

Editor (4 tools)

  • editor.play - Play current or custom scene
  • editor.stop - Stop playing
  • editor.pause - Toggle pause
  • editor.getState - Get playing state, version, project name

Debug (3 tools)

  • debug.screenshot - Capture viewport
  • debug.tree - Get scene tree as text
  • debug.log - Print message

Input (7 tools) - For Game Testing

  • input.keyPress - Press and release key {key: "W"}
  • input.keyDown - Hold key down
  • input.keyUp - Release key
  • input.mouseClick - Click at position {x, y, button: "left"|"right"|"middle"}
  • input.mouseMove - Move mouse to position {x, y}
  • input.actionPress - Press input action {action: "jump"}
  • input.actionRelease - Release input action

Script (2 tools)

  • script.list - List .gd files
  • script.read - Read script content

Resource (1 tool)

  • resource.list - List files by extension

Supported Keys for Input

A-Z, 0-9, SPACE, ENTER, ESCAPE, TAB, BACKSPACE, DELETE
UP, DOWN, LEFT, RIGHT
SHIFT, CTRL, ALT
F1-F12

Node Types for Creation

TypeDescription
Node2D2D spatial
Node3D3D spatial
Sprite2D2D sprite
CharacterBody2D2D character
CharacterBody3D3D character
RigidBody2D/3DPhysics body
Area2D/3DTrigger area
Camera2D/3DCamera
Label, ButtonUI elements

Tips

Input Simulation

  • Only works during Play mode
  • Use input.actionPress for mapped actions (from Input Map)
  • Use input.keyPress for direct key simulation

Finding Nodes

node.find {name: "Player"}      # By name substring
node.find {type: "Sprite2D"}    # By exact type
node.find {group: "enemies"}    # By group

Vector Properties

node.setProperty auto-converts dictionaries to Vector2/Vector3:

{path: "Cam", property: "zoom", value: {x: 2, y: 2}}  # → Vector2(2, 2)

Console Logs

console.getLogs {limit: 50}           # Last 50 lines
console.getLogs {type: "error"}       # Errors only
console.getLogs {type: "warning"}     # Warnings only

🔐 Security: Model Invocation Setting

When publishing to ClawHub, you can configure disableModelInvocation:

SettingAI Auto-InvokeUser Explicit Request
false (default)✅ Allowed✅ Allowed
true❌ Blocked✅ Allowed

Recommendation: true

Reason: During Godot development, it's useful for AI to autonomously perform supporting tasks like checking scene tree, taking screenshots, and inspecting nodes.

When to use true: For sensitive tools (payments, deletions, message sending, etc.)

As of v1.2.10 this skill ships with disableModelInvocation: true — it runs only on explicit user request.

Security & Privacy Disclosure

This skill drives a live Godot Editor. Full disclosure of capabilities and current limitations:

  • Local HTTP bridge has no authentication (current limitation): the MCP bridge (port 27183) accepts requests without auth or origin restriction. Keep it bound to localhost and never expose the port on shared or public networks — any local process could otherwise send editor commands. Token authentication and origin allow-listing are on the roadmap.
  • Destructive operations — confirm with the user before: modifying/deleting nodes, saving scenes, simulating input. Keep the project under version control before automation sessions.
  • Data visibility: tools can read and transmit project name, engine version, scene structure, script contents, and viewport screenshots to the connected agent. Don't use on projects whose contents must not leave the machine.
  • Trigger scope: routine-sounding requests ("clean up the scene", "save everything") map to state-changing editor operations — confirm once before the first state-changing call in a session.

License

Apache-2.0 — See LICENSE.md

Questions people ask

Does input simulation work outside Play mode?
No. The documented tools state that input simulation (keys, mouse, input actions) only works during Play mode.
Is there authentication on the local MCP bridge?
No. The document discloses that the MCP bridge on port 27183 accepts requests without authentication or origin restriction, so it should be kept bound to localhost and not exposed on shared networks.
Will an AI agent invoke these tools automatically?
No. As of v1.2.10 the skill ships with disableModelInvocation: true, so it runs only on explicit user request.

Related skills

Engine auto-update & discovery skill OpenClaw: scan, evaluate, update, test, install, verify, rollback — dengan validation gate dan safety agar update tidak merusak workspace.

Use when updating OpenClaw or debugging an OpenClaw instance after an update. This skill acts as a structured update runbook with emphasis on gateway startup...

21 installs2 stars

Record a screen demonstration and turn it into a reusable, parameterized OpenClaw SKILL.md.

Explain how to use OpenClaw in a local environment. Use when the user asks how to find OpenClaw config or workspace files, how to add or inspect skills, how...

16 installs

Route OpenClaw CLI requests to the right command family and profile, then verify the outcome.

336 installs7 stars