EasyOCR MCP tool

Name: EasyOCR MCP tool
Author: WindoC

@WindoC

代码仓库

关于 EasyOCR MCP tool

A Model Context Protocol (MCP) server that provides OCR capabilities using the EasyOCR library.

#ocr #gpu

基本信息

分类

其他

传输方式

stdio

发布者

WindoC

提交者

Antonio Cheong

配置

使用下面的配置,将此服务器添加到你的 MCP 客户端。

{
  "mcpServers": {
    "easyocr-mcp": {
      "command": "uv",
      "args": [
        "--directory",
        "/path/to/your/project/easyocr-mcp",
        "run",
        "easyocr-mcp.py"
      ],
      "env": {
        "EASYOCR_LANGUAGES": "en,ch_tra,ja"
      }
    }
  }
}

工具

未检测到工具

工具是从 README 中自动提取的。维护者可以在 ## Tools 标题下列出工具,即可填充这部分内容。

概览

What is EasyOCR MCP Server?

A Model Context Protocol (MCP) server that provides OCR capabilities using the open-source EasyOCR library. It allows AI agents to extract text from images (base64, files, or URLs) with support for over 80 languages.

How to use EasyOCR MCP Server?

Install dependencies with uv sync (optionally install PyTorch with GPU support). Configure language selection via the EASYOCR_LANGUAGES environment variable in your MCP client’s config.json. Run the server with uv run easyocr-mcp.py or mcp run easyocr-mcp.py, then invoke tools: ocr_image_base64, ocr_image_file, or ocr_image_url.

Key features of EasyOCR MCP Server

3 OCR Tools: base64, file, or URL input
Multi-language Support: 80+ languages with dynamic selection
Flexible Output: text-only or detailed results with coordinates and confidence
Performance Optimized: reader caching for better performance
Native EasyOCR Output: returns EasyOCR's original format

Use cases of EasyOCR MCP Server

Extracting text from scanned documents or photos
Reading text from images in multiple languages
Automating OCR from image URLs during web scraping
Processing base64-encoded images in chat or API workflows

FAQ from EasyOCR MCP Server

How do I select which languages to recognize?

Set the EASYOCR_LANGUAGES environment variable to a comma-separated list of language codes (e.g., en,ch_sim,ja). Default is English.

What output formats are available?

Set the detail parameter to 0 for text-only results (list of strings) or 1 for full details (list of tuples containing bounding box coordinates, text, and confidence score).

Does the server require a GPU?

GPU usage is determined at installation time based on your PyTorch installation. Install PyTorch with CUDA support to enable GPU acceleration; otherwise the server runs on CPU.

How do I process an image from a URL?

Use the ocr_image_url tool. The server will fetch the image from the URL and perform OCR automatically.

What are the threshold parameters for paragraph detection?

width_ths and height_ths control how close bounding boxes must be to be merged into a paragraph (default 0.7). The paragraph parameter enables paragraph detection (default false).