AnkleBreaker-Studio/unity-mcp-server

What it solves

AnkleBreaker Unity MCP connects AI assistants to the Unity game engine, allowing them to author scenes, manage game objects, run tests, and handle builds. It solves the difficulty of coordinating AI agents across multiple Unity projects and editors simultaneously, providing a structured way for agents to interact with the Unity Editor and running games without manual intervention.

How it works

The project consists of a Node.js server and a Unity plugin. The server implements the Model Context Protocol (MCP), acting as a bridge between AI clients (like LLM-based assistants) and the Unity Editor. It uses request-local routing to ensure that different agents can target specific Unity project instances. The Unity plugin handles the actual execution of commands within the editor, managing a queue to serialize writes and batch reads to maintain stability.

Who it’s for

Unity developers and technical artists who want to use AI agents to automate scene construction, iterate on multiplayer scenarios, perform profiling, or manage project-wide tasks across multiple editor instances.

Highlights

• Multi-Project Coordination: Supports independent request routing and separate queues for multiple open Unity editors.
• Extensive Toolset: Provides over 80 exposed MCP tools and hundreds of advanced tools for scene authoring, materials, profiling, and Unity Hub integration.
• Multiplayer Support: Includes specific tools for Multiplayer Play Mode (MPPM) and ParrelSync to coordinate host/client scenarios.
• Observability: Features a Unity Dashboard to monitor agent sessions, request queues, and action history with support for named undo.
• Performance Optimizations: Implements bounded compiler metadata reuse and paginated test results to reduce latency in code execution and data retrieval.