Funplay MCP for Unity

The Most Advanced MCP Server for Unity Editor

中文 | English

💖 If you find this project useful, please consider giving it a Star. It helps more Unity developers discover it and supports ongoing development.

Funplay MCP for Unity is an MIT-licensed Unity Editor MCP server that lets AI assistants like Claude Code, Cursor, Kimi Code, LM Studio, Windsurf, Codex, and VS Code Copilot operate directly inside your running Unity project.

Describe your game in one sentence — your AI assistant builds it in Unity through Funplay MCP for Unity's built-in tools for scene creation, script generation, runtime validation, input simulation, performance analysis, and editor automation.

"Build a snake game with a 10x10 grid, food spawning, score UI, and game-over screen"

Your AI assistant handles it through Funplay MCP for Unity: creates the scene, generates all scripts, sets up the UI, and configures the game logic — all from a single prompt.

16-second demo — AI generates a 3D model and integrates it into the scene end-to-end. Watch HD MP4.

Quick Start

If you just want to get connected fast, do these three things:

• Install the Unity package from the Git URL
• Start Funplay > MCP Server
• Use the built-in one-click client configuration

1. Install via UPM (Git URL)

In Unity, go to Window → Package Manager → + → Add package from git URL:

https://github.com/FunplayAI/funplay-unity-mcp.git

💡 Before you clone or install, a quick ⭐ on GitHub would be greatly appreciated.

Optional: Install via OpenUPM

If you want Unity Package Manager to show registry-backed package version history and allow version selection, install from OpenUPM instead of Git.

Using the OpenUPM CLI:

openupm add com.gamebooom.unity.mcp

Or add the scoped registry manually in Packages/manifest.json:

{
  "scopedRegistries": [
    {
      "name": "OpenUPM",
      "url": "https://package.openupm.com",
      "scopes": [
        "com.gamebooom"
      ]
    }
  ],
  "dependencies": {
    "com.gamebooom.unity.mcp": "0.6.9"
  }
}

If you installed from a Git URL before, remove the Git dependency first, then install from OpenUPM. Git-installed packages only show the resolved Git version in Unity and do not get the registry-backed Version History list.

2. Start the MCP Server

Menu: Funplay → MCP Server to start the server.

A new project gets its own port, derived from the project path, so two editors opened on different projects never fight over one port. The MCP Server window shows the exact URL (http://127.0.0.1:<port>/); the one-click configuration writes that URL for you. Type a port in Server Port to pin a fixed one (CI, a firewall rule) and Use Per-Project Port to derive one instead.

Upgrading from an earlier version changes nothing: the project keeps the port it was already using, recorded as a pin, so existing client configs keep working. Click Use Per-Project Port when you want that project to run beside another editor. See Running Several Unity Projects at Once for the full setup guide.

If the window reports a fallback port, do not use a client entry that still targets the occupied stable port; it could reach the process that owns that port. One-click configuration stays blocked until you click Use Per-Project Port or Pin Current Port and the server finishes restarting.

Direct in-process HTTP is the default transport. If you need stronger connection continuity across Unity script recompiles or Play Mode domain reloads, enable Experimental Broker Mode in the MCP Server window. It runs a tiny local broker with Unity's bundled Mono, keeps the same 127.0.0.1 port for MCP clients, and requires no client config change.

Open Funplay → Tool Exposure if you want to edit the exact tools exposed by core or full.

Open Funplay → MCP Settings if you need to adjust execute_code safety defaults or plugin debug logging.

3. Configure Your AI Client

Use the built-in One-Click MCP Configuration in the Funplay > MCP Server window first.

Select your target client, click Configure, and the package writes the recommended MCP config entry for you.

For Claude Code, Cursor, Codex, OpenCode, and DeepSeek Harness, click Configure + Skills to also install both built-in project skills.

If you want project-specific AI guidance for the current Unity project, open Funplay → Project Skills to choose supported platforms and install the built-in unity-mcp-workflow and unity-ui-composition skills. The UI skill covers responsive portrait and landscape uGUI work.

If you prefer to edit config files manually, use the examples below as fallback references. Replace <project> with this project's entry name and <port> with its port -- the MCP Server window shows both:

{
  "mcpServers": {
    "funplay-<project>": {
      "type": "http",
      "url": "http://127.0.0.1:<port>/"
    }
  }
}

{
  "mcpServers": {
    "funplay-<project>": {
      "url": "http://127.0.0.1:<port>/"
    }
  }
}

{
  "mcpServers": {
    "funplay-<project>": {
      "url": "http://127.0.0.1:<port>/"
    }
  }
}

{
  "servers": {
    "funplay-<project>": {
      "type": "http",
      "url": "http://127.0.0.1:<port>/"
    }
  }
}

{
  "mcpServers": {
    "funplay-<project>": {
      "url": "http://127.0.0.1:<port>/"
    }
  }
}

{
  "mcpServers": {
    "funplay-<project>": {
      "type": "http",
      "url": "http://127.0.0.1:<port>/"
    }
  }
}

{
  "mcpServers": {
    "funplay-<project>": {
      "url": "http://127.0.0.1:<port>/"
    }
  }
}

[mcp_servers.funplay-<project>]
url = "http://127.0.0.1:<port>/"

{
  "mcp": {
    "funplay-<project>": {
      "type": "remote",
      "url": "http://127.0.0.1:<port>/",
      "enabled": true
    }
  }
}

# >>> funplay-mcp:funplay-<project> begin (managed by Funplay MCP -- reconfigure from Unity > Funplay > MCP Server)
# Funplay Unity MCP endpoint served by this editor; tools appear as mcp__funplay-<project>__<tool>.
- insert:
    - id: mcp-funplay-<project>
      name: '@deepseek-ai/dsh-mcp-client'
      config:
        serverName: funplay-<project>
        transport: streamable-http
        url: http://127.0.0.1:<port>/
# <<< funplay-mcp:funplay-<project> end

{
  "mcpServers": {
    "funplay-<project>": {
      "serverUrl": "http://127.0.0.1:<port>/"
    }
  }
}

4. Verify the Connection

Open your AI client and try a few safe requests first:

• "Call get_scene_info and tell me what scene is open."
• "Read unity://project/context and summarize the current editor state."
• "Use execute_code to return the active scene name."

If those work, the MCP server, resources, and primary execution tool are connected correctly.

5. Start Building

Open your AI client and try: "Create a 3D platformer level with 5 floating platforms"

Before You Start

• This package is Editor-only. It does not add runtime components to your built game.
• The MCP server port is derived per project (range 20000-29999) for new projects, or pinned — projects upgraded from an earlier version keep their existing port as a pin, and any port you type is a pin. The MCP Server window shows where the port came from and the active URL. A project's client-config entry is named after the project directory (for example funplay-love-town), so configuring several projects no longer overwrites one shared funplay entry. Two projects that share a product name would resolve to the same entry name; the second one configured appends a project hash automatically, so nothing is overwritten and no setting has to be turned on.
• Local MCP server settings are stored in UserSettings/FunplayMcpSettings.json.
• The package defaults to core: 40 focused tools for structured inspection/editing, UI audits and authoring, recoverable Editor preparation, shared task status, visual evidence, input and logs. Short MCP tasks can complete in one response; get_task supports bounded waits and state-change revisions. execute_code remains a project-specific fallback. full retains all 180 tools, including legacy status/compile/Play APIs, configuration, preview management and specialized diagnostics. Custom exposure lists are preserved; get_tool_capabilities distinguishes implementation from exposure.
• execute_code safety checks and the stricter filesystem guard are enabled by default from Funplay > MCP Settings. The guard blocks obvious destructive snippets, broad System.IO writes, raw file streams, and absolute/user/system/traversal paths, but it is not a complete sandbox. Clients may still override the default per call with the optional safety_checks argument.
• Plugin debug logging is off by default and can also be enabled from Funplay > MCP Settings. Warnings and errors are always written to the Unity Console.
• All exposed MCP tools run directly. There is no extra approval toggle.
• Menu: Funplay > Check for Updates can refresh Git installs in place or download and import the latest unitypackage automatically.

Why This Project

• execute_code First — Optimized around one in-memory C# execution tool for rich editor/runtime orchestration. See execute_code: In-Memory C# Execution below for details.
• Default Safety Checks — execute_code now has persistent default-on safety toggles, including a stricter filesystem guard for clients that do not expose per-call arguments clearly
• Play Mode Automation — Enter play mode, simulate keyboard/mouse input, capture screenshots, inspect logs, and validate behavior from the same MCP session
• Project Context Built In — Exposes live resources for project state, active scene, selection, compilation, console output, and MCP interaction history
• Focused by Default, Full When Needed — core exposes 40 focused tools; full exposes all 180 built-in tools
• Single Unity Package — No extra approval UI, no external daemon to click through, and no Python requirement for the Unity-side plugin itself
• Extensible — Add custom tools with attribute-based discovery, or connect Unity to external MCP services when needed

Highlights

• 180 Built-in Tools — Scene editing, assets, scripts, play mode control, screenshots, performance analysis, prompts, resources, structured object location, SerializedObject-based component editing, editor-state inspection, menu-item fallback, and editor automation across 42 modules
• Structured Returns + instanceId Chaining — Tools return {success, message, data} JSON with stable instanceId fields so agents can chain by_id calls reliably instead of re-resolving by name
• IFunplayCommand for execute_code — New snippet template with auto-Undo (ctx.RegisterObjectCreation/Modification/DestroyObject), structured logs (ctx.Log/LogWarning/LogError), and a tracked changelog returned to the agent
• Resources & Prompts — Live project context, scene/selection/error resources, resource templates, and reusable workflow prompts
• Input Simulation + Screenshots — Drive play mode with keyboard/mouse simulation and verify results with game/scene captures
• Built-in Updating — Check for updates from the Unity menu and either re-pull the Git package or auto-import the latest unitypackage
• One-Click Client Configuration — Generate MCP config entries for Claude Code, Cursor, Kimi, LM Studio, VS Code, Kiro, Trae, Codex, OpenCode, DeepSeek Harness, Antigravity, and similar clients directly from the Unity window
• Tool Exposure Control — Edit the exact tools exposed by core and full
• Project Skills Manager — Configure project-level skills for supported AI clients, with built-in unity-mcp-workflow and unity-ui-composition guidance
• MCP Settings — Adjust execute_code safety defaults and enable verbose plugin debug logging when troubleshooting MCP connections or tool execution
• Vendor Agnostic — Works with any AI client that supports MCP: Claude Code, Cursor, Kimi Code, LM Studio, Windsurf, Codex, VS Code Copilot, etc.

execute_code: In-Memory C# Execution

execute_code is the heart of Funplay MCP for Unity. It lets an AI write a C# snippet, compile it through a Roslyn-first in-memory flow, and run it on the editor thread — the agent gets the full Unity Editor and runtime API surface without writing any project files to disk.

• Zero project footprint compilation — Snippets are compiled with Unity's bundled Roslyn csc first while preserving the in-memory compilation/execution flow. No .cs files are written under Assets/, no domain reload is triggered, no project state is touched beyond what the snippet itself does.
• Editor-ready before it runs — Each call refreshes the AssetDatabase and waits for any pending compilation to settle before compiling the snippet, so external file edits are picked up automatically without a separate request_recompile.
• Auto-Undo + structured logs (recommended template) — Implement IFunplayCommand and use the injected ExecutionContext so every created / modified / destroyed object participates in editor Undo, and the changelog is returned to the agent.

using UnityEngine;
using UnityEditor;
using Funplay.Editor.Tools.Helpers;
using Funplay.Editor.Tools.Scripting;

public class CommandScript : IFunplayCommand
{
    public void Execute(ExecutionContext ctx)
    {
        var go = GameObject.CreatePrimitive(PrimitiveType.Cube);
        ctx.RegisterObjectCreation(go);          // auto-Undo + tracked
        ctx.Log("Created {0}", go.name);
        ctx.ReturnValue = GameObjectSerializer.Describe(go, includeComponents: false);
    }
}

The response carries { logs, created, modified, destroyed, returnValue }, so the agent can verify exactly what changed without re-querying the scene.

The legacy template (public static string Run()) is still supported — useful for one-off inspection snippets where structured tracking is overkill.

When to reach for execute_code vs a specialized tool — execute_code shines for multi-step orchestration, novel reads, and situations where chaining 5–10 narrow tool calls would be noisier than one snippet. For single-field component edits, simple selection changes, or anything covered by an existing tool, prefer the dedicated tool — it is cheaper for the LLM to call and easier to verify.

Comparison With Coplay

The table below compares this repository with the publicly documented behavior of Coplay's open-source unity-mcp repository on GitHub.

Source for Coplay column: CoplayDev/unity-mcp

Comparison With Unity AI Assistant

The table below compares this repository with Unity Technologies' official com.unity.ai.assistant package (v2.7.0-pre.2 as of 2026-05).

For a long-form comparison of the two approaches see Funplay Unity MCP vs Unity AI Assistant detailed comparison (Chinese).

MCP Capabilities

The current open-source package exposes four high-value capability layers:

• Tools — 180 total tools in full, 40 focused tools in core
• Primary execution — execute_code for rich editor/runtime orchestration
• Prompts — parameterized workflow prompts: edit_prefab_safely, verify_compilation, enter_play_and_recover, wire_serialized_references, create_playable_prototype. Projects can add their own through mcp-prompts/*.md files in the project root.
• Resources — project context, scene summaries, selection state, compile errors, console errors, MCP interaction history, plus resource templates for scene objects, components, and asset paths

Project Prompts

Project prompt files use a small, dependency-free front-matter format followed by the workflow body:

---
name: validate_activity
description: Open and validate a project activity.
arguments: activity_key(required), theme_id
---
Open activity {activity_key} with theme {theme_id}, then validate its runtime state.

Names and argument names must match [a-z][a-z0-9_-]{0,63}. Required, unknown, and non-string arguments are rejected by prompts/get; omitted optional placeholders become empty strings. Definitions are cached when the MCP server starts, so restart the server or trigger a Unity domain reload after changing a prompt file.

Built-in Tools

Version 0.6.9 includes 180 tool functions across 42 modules, with 40 high-frequency tools exposed by default.

📊 See PROFILER_TOOLS.md for the full Profiler tool reference, implementation notes, known limitations, and test report.

See Reliable UI workflows for end-to-end preparation, read-only audits, structured queries, project TMP/prefab/input defaults, capture-coordinate mapping, recording markers/keyframes, and recoverable preview sessions. Both built-in Project Skills include these workflows; use Project Skills to update existing installations. These tools and skill updates are included starting with v0.6.9.

Recording Game View video

Use record_game_view (included in core) to review UI animations, transitions, and multi-step interactions in a short silent MP4. It uses Unity's built-in MediaEncoder; no Recorder package or external encoder is required.

1. Enter Play Mode, wait for reload recovery, and open the Game tab so it renders.
2. Call record_game_view with {"action":"start","duration_seconds":10,"fps":15,"max_dimension":1280}. Keep its recording_id.
3. Perform the interactions while recording. Start returns immediately, and the duration limit stops recording automatically.
4. Poll with {"action":"status","recording_id":"<id>"}. To end early, use action:"stop", then poll until finalization finishes.
5. When ready:true, read the returned path as a local video file. Video bytes are not embedded in the MCP response; remote clients need separate access to the Unity machine's files.

Clips receive unique filenames under <UnityProject>/Library/FunplayMcp/Recordings/ and are retained until you remove them. Start accepts 1–120 seconds, 1–60 target FPS, and a maximum dimension of 128–1920 pixels; it preserves the aspect ratio without upscaling and uses even encoding dimensions. State includes actual frame count, elapsed time, stop reason, file size, and errors. Pass the recording ID when polling/stopping to avoid accidentally controlling a newer clip.

Currently supports graphics-enabled macOS and Windows Editors in Play Mode, with no audio track. It captures the rendered Game View, including overlay UI, without re-rendering scene cameras. Keep the Game tab rendering and do not resize it during capture: hiding/closing the tab or changing its render resolution stops recording with an error instead of silently capturing stale or stretched frames. Exiting Play Mode finalizes the clip; a script/domain reload finalizes it early and preserves an interrupted receipt in the editor session. Only consume a file with ready:true. Slow frames keep their real timestamps, but recording adds overhead: this is visual evidence, not frame-accurate performance profiling.

Adding Custom Tools

Create your own tools with simple attribute annotations:

using System.ComponentModel;

[ToolProvider("MyTools")]
public static class MyCustomTools
{
    [Description("Spawns enemies at random positions in the scene")]
    public static string SpawnEnemies(
        [ToolParam("Number of enemies to spawn", Required = true)] int count,
        [ToolParam("Prefab path in Assets")] string prefabPath)
    {
        // Your implementation here
        return $"Spawned {count} enemies";
    }
}

Methods are automatically discovered, converted to snake_case (spawn_enemies), and exposed via MCP with JSON Schema definitions.

Architecture

MCP Server (HTTP JSON-RPC 2.0)
    └─ MCPRequestHandler (protocol handling)
        └─ MCPExecutionBridge
            └─ FunctionInvokerController (reflection-based invocation)
                └─ Tool Functions (180 built-in tools across 42 modules)

External AI Client → HTTP Request → MCPRequestHandler → MCPExecutionBridge → FunctionInvokerController → tool method

Requirements

• Unity 2022.3 or later
• .NET / Mono with Newtonsoft.Json

Contributing

Contributions are welcome! Please read the Contributing Guide before submitting a PR.

License

MIT — Free to use, modify, distribute, and integrate into commercial or open-source projects.