real-eye
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@real-eyeread my current window and tell me where the Send button is"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
真实之眼 Real Eye
给 AI 一双能看懂 Windows 界面的眼睛。
它把任意窗口读成带坐标的文字——控件的名字、角色、能不能点,加上画面上所有文字和它们的位置。于是:
看不见屏幕的模型也能操作电脑:读出来的是文字,不是截图,纯文本模型(DeepSeek、Kimi、长上下文那些)同样能用
点击不用猜像素:知道按钮叫什么、在哪,点哪儿是算出来的
被遮挡也能读:抓的是窗口自己渲染的画面,窗口在后面、被别的窗口盖住都行
它是一个 MCP 服务,所以任何支持 MCP 的客户端都能用:Claude Code、Cursor、Cline、Windsurf……
只读:只看,不点、不打字、不抢焦点、不激活窗口。
装
需要 Windows 10/11 和 Python 3.9+。
git clone <这个仓库>
cd real-eye
pip install -r requirements.txt默认用 Windows 自带 OCR,装完约 30MB。想要更准的中文识别,把 requirements.txt 里 rapidocr_onnxruntime 那行取消注释再装一次(多约 240MB)。
Related MCP server: Little Mouse Pointer
配
在 MCP 客户端里加一段(以 Claude Code 的 ~/.claude.json 为例):
{
"mcpServers": {
"real-eye": {
"command": "python",
"args": ["C:\\路径\\到\\real-eye\\real_eye_server.py"]
}
}
}Windows 上建议写 python.exe 的绝对路径。
重启客户端即可。模型就能调用下面这些工具了。
工具
工具 | 做什么 |
| 列出当前可见的窗口 |
| 读一个窗口:控件清单(角色/名字/坐标/可点/启用状态)+ 画面上所有文字和坐标。不填目标就读当前前台窗口。 |
| 按名字找东西在哪:优先查控件表(精确),没有再找 OCR 文字。返回屏幕坐标 |
| 等到某段文字出现(或超时)。用于等界面反应,比如等"发送"亮起来 |
eye_see / eye_find 有个 accurate 开关:默认快(Windows 自带 OCR,约 0.4 秒),设为 true 换 RapidOCR(约 4 秒,中文更准)。
快不快
本机实测(900×600 的窗口):
操作 | 耗时 |
找控件(名字在控件表里) | 0.2 秒 |
读一次窗口(系统 OCR) | 0.4 秒 |
重复读同一画面(画面没变) | 0.1 秒 |
读一次窗口( | 3–5 秒 |
已知限制
只读。要点击/输入请配一个 Computer Use 类工具,用这里返回的坐标去点。
系统 OCR 会认错个别汉字("询问"→"词问")。要紧的文字加
accurate=true。有些窗口既没有控件表、又不肯自己渲染(个别游戏、某些硬件加速界面),这时会明确报错,不会返回错的内容。
目标窗口最小化时拒绝读取。
许可
MIT,见 LICENSE。用到的第三方库各有自己的许可证(pywin32 PSF、mss MIT、Pillow MIT-CMU、numpy BSD、uiautomation Apache-2.0、winsdk MIT、rapidocr_onnxruntime Apache-2.0)。
This server cannot be deployed
Maintenance
Related MCP Connectors
Eyes and hands on real Windows PCs — observe, click, type via Glasswarp API.
OCR, transcription, file extraction, and image generation for AI agents via MCP.
Give agents eyes on any web page: structured context, and changes explained in plain language.
Web scraping for agents. Point it at a URL and it returns the page as clean markdown, JavaScript-rendered pages included. Point it at a site and it maps the URLs or crawls the section you need in the background, a few pages at a time so results fit in the conversation. Search the web and read full pages, extract fields with a JSON schema you define (validated, never invented), read a store's catalogue or a blog's posts from the platform's own feed, and check whether a page has changed. Failed requests cost nothing. The free plan includes 1,500 credits a month.
Related MCP Servers
- AlicenseAqualityCmaintenanceAllows AI clients to see and control Windows 10/11 desktops via MCP, with screenshots, UI Automation, Chrome CDP, keyboard/mouse, and terminal using semantic element targeting.30649 npmMIT
- AlicenseNot gradedqualityBmaintenanceEnables AI agents to select Windows desktop UI elements, windows, or screen regions via mouse hover, then obtain context through UI Automation, screenshots, and local OCR for MCP-compatible clients.2MIT
- AlicenseNot gradedqualityBmaintenanceEnables reading and interacting with native Windows application UI through the UI Automation API, allowing listing windows, extracting structured content, and performing clicks/typing without screenshot-based vision.2MIT
- AlicenseNot gradedqualityBmaintenanceEnables AI clients on Windows to control the local mouse, keyboard, and screen understanding via MCP stdio, allowing automated workflows like viewing the screen, locating elements, clicking, typing, and verifying results.1Apache 2.0