AnkleBreaker-Studio/unity-mcp-server

What it solves

This project provides a comprehensive bridge between AI assistants (like Claude, Cursor, and Windsurf) and the Unity game engine. It allows developers to control the Unity Editor and Unity Hub directly from an AI chat interface, eliminating the need to manually perform repetitive editor tasks, scene setup, and project management.

How it works

It implements the Model Context Protocol (MCP), acting as a server that translates AI commands into actions within Unity. The architecture consists of three parts: the AI assistant, the MCP server (this repository), and a companion Unity Editor plugin that runs an HTTP API bridge. The server communicates with the Unity Hub CLI for installation tasks and the Unity plugin for real-time editor manipulation.

Who it’s for

Unity game developers who want to automate their workflow, rapidly prototype scenes, or use AI as a co-pilot for game development tasks ranging from simple object placement to complex shader and terrain editing.

Highlights

• Massive Toolset: Over 330 tools across 30+ categories, including scene management, physics, lighting, animation, and profiling.
• Deep Integration: Supports advanced Unity features like Shader Graph, Amplify Shader Editor, NavMesh baking, and the Memory Profiler.
• Intelligent Tool Management: Uses a two-tier system (Core and Advanced tools) to prevent overwhelming the AI assistant's context window.
• Robust Connectivity: Features multi-instance discovery, port resilience, and crash detection to maintain stable connections during editor restarts or compiles.
• Visual Feedback: Ability to capture screenshots of the scene and game views for the AI to visually inspect the result of its actions.