MCP Point Reader

Expose [Point Reader] features to AI assistants (Claude Desktop, etc.) through the MCP protocol, enabling:

• Screenshot OCR
• Clipboard text capture
• Text-to-speech (offline / online TTS)
• Vocabulary management (add, list, search, mark learned, delete)
• Reading history query

Prerequisites

• Python 3.10+
• [Point Reader] installed with config.json and data.db generated

Installation

cd mcp-point-reader
pip install -e .

Dependencies: mcp, PyQt5, edge-tts, pyttsx3, pyperclip, pytesseract, Pillow, pynput

Configuration

The MCP plugin reads Point Reader's config automatically. The default path is D:\English.

If Point Reader is installed elsewhere, set the environment variable:

set POINT_READER_DIR=D:\your-path\English

Usage with Claude Desktop

Edit claude_desktop_config.json (located at %APPDATA%\Claude\):

{
  "mcpServers": {
    "point-reader": {
      "command": "python",
      "args": ["-m", "server"],
      "cwd": "D:\\mcp-point-reader"
    }
  }
}

Replace cwd with the actual project path.

Restart Claude Desktop. The AI can now use Point Reader's features in conversations.

Available Tools

| Tool | Description | | ------------------- | ----------------------------------------------------- | | screenshot_ocr | Full screenshot + OCR text recognition | | capture_clipboard | Read clipboard text | | speak_text | Speak text aloud (auto-selects Chinese/English voice) | | add_vocab | Add a word to vocabulary | | list_vocab | List vocabulary (with filtering / search) | | search_vocab | Search vocabulary by keyword | | mark_learned | Mark a word as learned | | delete_vocab | Delete a word from vocabulary | | get_history | View reading history |

Manual Test

python -m server

The server will wait for JSON-RPC messages from an MCP client. Claude Desktop connects automatically — no manual interaction needed.

License

MIT

MCP Point Reader

通过 MCP 协议向 AI 助手（Claude Desktop 等）暴露[点读助手]的功能，让 AI 可以：

• 截图识别文字（OCR）
• 读取剪贴板内容
• 朗读文本（离线 / 在线 TTS）
• 管理生词本（增删改查、标记已掌握）
• 查看朗读历史

前置条件

• Python 3.10+
• [点读助手]已安装，并已生成 config.json 和 data.db

安装

# 克隆或进入项目目录
cd mcp-point-reader

# 安装依赖
pip install -e .

国内网络慢可加镜像：pip install -e . -i https://mirrors.aliyun.com/pypi/simple/

依赖包括：mcp, PyQt5, edge-tts, pyttsx3, pyperclip, pytesseract, Pillow, pynput

配置

MCP 插件会自动读取点读助手的配置文件，默认路径为 D:\English\。

如果点读助手安装在其他位置，设置环境变量：

set POINT_READER_DIR=D:\your-path\English

在 Claude Desktop 中使用

编辑 claude_desktop_config.json（位于 %APPDATA%\Claude\）：

{
  "mcpServers": {
    "point-reader": {
      "command": "python",
      "args": ["-m", "server"],
      "cwd": "D:\\mcp-point-reader"
    }
  }
}

cwd 替换为实际的项目所在路径。

保存后重启 Claude Desktop，即可在对话中让 AI 使用点读助手的功能。

可用工具

| 工具 | 说明 | | ------------------- | ---------------------------------- | | screenshot_ocr | 全屏截图并识别文字 | | capture_clipboard | 读取剪贴板文本 | | speak_text | 朗读指定文本（自动选择中英文语音） | | add_vocab | 添加单词到生词本 | | list_vocab | 查看生词本（支持筛选/搜索） | | search_vocab | 搜索生词本中的单词 | | mark_learned | 标记单词为已掌握 | | delete_vocab | 从生词本中删除单词 | | get_history | 查看朗读历史 |

手动测试

python -m server

启动后程序会等待 MCP 客户端的 JSON-RPC 消息。可配合 Claude Desktop 自动连接，无需手动操作。

许可证

MIT