Skip to main content

Overview

MCP for Unity bridges AI assistants — Claude, Codex, VS Code, local LLMs, and more — with the Unity Editor via the Model Context Protocol. Give your LLM the tools to manage assets, control scenes, edit scripts, run tests, and automate workflows.

MCP for Unity building a scene

What you get​

  • 40+ Unity Editor tools exposed over MCP — manage_scene, manage_script, manage_gameobject, manage_material, manage_physics, run_tests, and more.
  • 25+ read-only resources for state introspection — editor_state, gameobject_components, project_info, unity_instances, etc.
  • Auto-configuration for popular MCP clients — Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Cline, Codex, Qwen, Gemini CLI, Copilot CLI, OpenClaw.
  • Multi-instance support — drive several Unity Editors from a single session via set_active_instance.
  • Two transports — HTTP (multi-agent, default) and stdio (single-agent legacy).

When you'd use it​

  • Prototype scenes and gameplay with natural language ("build a player controller with WASD and a double-jump").
  • Generate and refactor C# scripts with full project context and validation.
  • Automate repetitive editor tasks — bulk asset processing, scene validation, regression testing.
  • Build custom AI-driven editor tools on top of the MCP protocol.

Next steps​

  • Install — Add the Unity package, install the Python server, and connect your first MCP client.
  • Your First Prompt — End-to-end "build me a red cube" tutorial.
  • Choosing an MCP Client — A capability matrix across all supported clients.
  • Setup Wizard (coming soon) — Walk through the first-run experience.

MIT licensed. Sponsored and maintained by Aura. Not affiliated with Unity Technologies.