macOS OCR

Provides offline OCR using macOS Vision with JSON/BBOS-layout outputs for LLM-based document reconstruction.

python

1

GitHub Stars

python

Language

5 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "wenjiazhu-macos-ocr-mcp": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/wenjiazhu/macos-ocr-mcp.git",
        "macos-ocr"
      ]
    }
  }
}

You can run a macOS-native OCR MCP server locally to perform offline, high-precision OCR using macOS Vision. It provides multi-language recognition, PDF support, smart block aggregation, structured JSON output for LLM workflows, strong privacy since everything runs on your machine, and zero-configuration startup via UVX.

How to use

You integrate this MCP server with your MCP client to perform OCR tasks on images and PDFs. Use the MCP server to either extract plain text or reconstruct layout information for subsequent document rebuilding. For text-focused tasks, prefer the read_image_text tool. For layout-aware outputs suitable for keeping formatting, prefer the read_image_layout tool. The outputs are designed to be composable with your workflows, including JSON blocks with coordinates and lines for precise reassembly.

How to install

Prerequisites you need before starting:

macOS 10.15+ (Catalina) or newer
Python 3.10+
Homebrew installed on your Mac

Install the MCP runtime tool UV via Homebrew, which enables zero-config startup for MCP servers.

Install UV with Homebrew:

brew install uv

To run the MCP server locally, you can use the UVX workflow described in the configuration example. The following is the configuration you would place in your client’s MCP configuration for Claude Desktop integration.

{
  "mcpServers": {
    "macos-ocr": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/wenjiazhu/macos-ocr-mcp.git",
        "macos-ocr"
      ]
    }
  }
}

Available tools

read_image_text

Recognizes text from an image or PDF, performing paragraph merging and table optimization to output plain text.

read_image_layout

Extracts structured layout information as blocks with text, bbox coordinates, and underlying lines to enable layout-aware reconstruction and formatting for downstream processing.