playcaller
An MCP server that enables AI agents to play-test Unity games by capturing screenshots and simulating inputs like taps, drags, and key presses, acting as a Playwright for Unity.
README
Playcaller
An MCP server that lets AI agents play-test Unity games like a real player. It provides screenshot capture, input simulation (tap, drag, flick, key press), Play Mode control, and more — all accessible from AI agents.
In short, it's a Playwright MCP for Unity.
Features
- Game View screenshot capture
- Tap, drag, flick, and key input simulation
- Play Mode start / stop / pause
- Scene GameObject hierarchy inspection
- Console log retrieval
- AssetDatabase refresh
- Menu item execution
- Editor state retrieval
Limitations
- Not suited for testing fast-paced action games
- Not suited for evaluating overall game experience
- Not suited for judging visual effects quality
Requirements
- Unity 2022.3 or later
- Python 3.10 or later (uv recommended)
Setup
1. Install the Unity package
- Open Window > Package Manager in the Unity Editor menu bar
- Click the + button in the top-left corner
- Select Add package from git URL...
- Enter the following URL and click Add
https://github.com/takashicompany/playcaller.git
Newtonsoft.Json (com.unity.nuget.newtonsoft-json) will be installed automatically if not already present.
2. Register the MCP server
Run the following command:
claude mcp add playcaller -- uvx playcaller
If something doesn't work, try asking the AI for help.
3. Verify the connection
Launch Claude Code while the Unity Editor is open, and it will connect automatically.
Updating
Unity package
- Open Window > Package Manager
- Select Playcaller from the list
- Click Update
MCP server (PyPI)
uvx --upgrade playcaller
and restart Claude Code.
Notes
Run In Background
If the game does not progress after starting Play Mode via MCP, the project's "Run In Background" setting may be disabled. Enable it in Edit > Project Settings > Player > Resolution and Presentation > Run In Background, or set Application.runInBackground = true in your game code.
HDR projects
In URP/HDRP projects with HDR enabled, playcaller_screenshot may hang due to a known Unity issue. Use playcaller_read_gameview_pixels instead, which reads the GameView's internal buffer directly and works with HDR.
MCP Tools
| Tool | Description |
|---|---|
playcaller_screenshot |
Capture a Game View screenshot |
playcaller_read_gameview_pixels |
Capture Game View by reading internal buffer (HDR compatible) |
playcaller_tap |
Tap at specified coordinates |
playcaller_multi_tap |
Execute multiple taps in sequence |
playcaller_drag |
Drag operation |
playcaller_flick |
Flick (quick swipe) operation |
playcaller_key_press |
Keyboard input |
playcaller_playmode |
Start / stop / pause Play Mode |
playcaller_wait |
Wait for a specified time or frame count |
playcaller_console_log |
Retrieve Unity console logs |
playcaller_refresh |
Refresh AssetDatabase |
playcaller_get_hierarchy |
Get the GameObject hierarchy in the scene |
playcaller_get_gameobject |
Get detailed info about a specific GameObject |
playcaller_execute_menu_item |
Execute a Unity Editor menu item |
playcaller_get_editor_state |
Get the current editor state |
playcaller_game_query |
Send a custom query to the running game |
playcaller_set_game_view_size |
Set the Game View resolution |
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。