Model Context Protocol moderate risk

Unity

Unity MCP enables AI-powered game development by connecting LLMs to Unity for runtime AI, code assistance, debugging, and workflow automation.

Connections & Capabilities

Connects To

GitHubDiscordDocker

Capabilities

readwriteexecadmin

Quickstart

Config

{
  "mcpServers": {
    "ai-game-developer": {
      "url": "http://localhost:8080"
    }
  }
}

Exposed MCP Tools (43)

moderate

copy

Copies an asset from one path to another.

Can overwrite existing assets if not used carefully.

moderate

create-folder

Creates a new folder in the project.

Can create unnecessary folders, cluttering the project.

critical

delete

Deletes assets from the project.

Permanent data loss if used incorrectly.

safe

find

Searches the asset database.

Read-only operation.

safe

get-data

Gets asset data, including serializable fields and properties.

Read-only operation.

high

modify

Modifies asset files in the project.

Can corrupt asset data if not used carefully.

moderate

move

Moves assets within the project.

Can break references if not used carefully.

safe

refresh

Refreshes the AssetDatabase.

Read-only operation.

moderate

create

Creates a new material asset.

Non-destructive write operation.

safe

list-all

Lists all available shaders in the project.

Read-only operation.

moderate

instantiate

Instantiates a prefab in the current scene.

Adds objects to the scene.

safe

open

Opens a prefab for editing.

Does not modify the prefab itself.

safe

close

Closes the currently opened prefab.

Does not modify the prefab itself.

moderate

save

Saves a prefab in editing mode.

Modifies the prefab asset.

critical

destroy

Destroys a GameObject and all its children.

Permanent data loss.

moderate

duplicate

Duplicates a GameObject.

Adds new objects to the scene.

moderate

set-parent

Sets the parent of a GameObject.

Can alter the scene hierarchy.

moderate

add

Adds a component to a GameObject.

Modifies the GameObject's properties.

safe

get

Gets information about a component on a GameObject.

Read-only operation.

safe

list

Lists available components.

Read-only operation.

moderate

create

Creates a new scene.

Creates a new asset.

safe

get-data

Retrieves the list of root GameObjects in a scene.

Read-only operation.

safe

list-opened

Lists currently opened scenes.

Read-only operation.

safe

open

Opens a scene.

Does not modify the scene file.

moderate

save

Saves the opened scene.

Modifies the scene file.

safe

set-active

Sets the active scene.

Does not modify any assets.

safe

unload

Unloads a scene.

Does not modify any assets.

critical

delete

Deletes script files.

Permanent data loss.

critical

execute

Compiles and executes C# code dynamically.

Arbitrary code execution.

safe

read

Reads the content of a script file.

Read-only operation.

high

update-or-create

Updates or creates a script file.

Can overwrite or create malicious scripts.

safe

get-logs

Retrieves Unity Editor logs.

Read-only operation.

safe

get-state

Returns information about the Unity Editor application state.

Read-only operation.

moderate

set-state

Controls the Unity Editor application state.

Can disrupt the editor's operation.

safe

get

Gets information about the current selection in the Unity Editor.

Read-only operation.

safe

set

Sets the current selection in the Unity Editor.

Does not modify any assets.

high

method-find

Finds a method using C# reflection.

Can expose private methods.

critical

method-call

Calls any C# method with input parameters.

Can execute arbitrary code.

safe

run

Executes Unity tests.

Does not modify any assets.

moderate

add

Installs a package from the Unity Package Manager.

Adds new code to the project.

moderate

remove

Removes a package from the Unity project.

Removes code from the project.

safe

list

Lists all installed packages.

Read-only operation.

safe

search

Searches for packages.

Read-only operation.

Safety Assessment

Unity MCP provides powerful tools for AI-assisted game development, but its capabilities also introduce significant risks. It is relatively safe when used for read-only operations and code generation in a controlled environment. However, the ability to modify and delete assets, execute arbitrary code, and access internal application state makes it risky if not properly secured and monitored.

Access to a wide range of LLM providers allows for flexibility and potential cost optimization.
Extensible architecture allows for the creation of custom MCP Tools, tailoring functionality to specific needs.
Offers debugging support, potentially reducing errors.
Provides tools for managing assets, game objects, and scenes in a controlled manner.

Includes tools that can modify and delete assets, game objects, and scripts, posing a risk of data loss or corruption.
The `Script/execute` tool allows for dynamic code execution, which can introduce security vulnerabilities if not carefully managed.
Access to reflection tools (`method-find`, `method-call`) can bypass normal access restrictions and potentially destabilize the application.
The ability to modify game objects and components at runtime introduces the risk of unexpected behavior or exploits.
Lack of clear sandboxing for script execution and asset modification.