Extractous MCP Server

A Model Context Protocol (MCP) server for text extraction using the extractous library.

Features

• Extract text from various file formats (PDF, DOCX, HTML, images with OCR, audio/video with speech-to-text)
• Get file metadata along with extracted text
• Support for MIME type override
• List supported file formats

Installation

# Install dependencies
uv pip install -e .

# Or using pip
pip install extractous-mcp

Usage

As MCP Server

# Run the server
extractous-mcp

# Or run directly
python -m extractous_mcp.server

Available Tools

1. extract_text

   • Extract text from a file
   • Parameters: file_path (required), mime_type (optional)

2. extract_text_with_metadata

   • Extract text and metadata from a file
   • Parameters: file_path (required), mime_type (optional)

3. get_supported_formats

   • Get list of supported file formats
   • No parameters

Example Usage

# Extract text from PDF
extract_text("/path/to/document.pdf")

# Extract with specific MIME type
extract_text("/path/to/file", mime_type="application/pdf")

# Get text with metadata
extract_text_with_metadata("/path/to/document.docx")

# Check supported formats
get_supported_formats()

Supported Formats

• Documents: PDF, DOC, DOCX, RTF, ODT
• Spreadsheets: XLS, XLSX, ODS, CSV
• Presentations: PPT, PPTX, ODP
• Web: HTML, HTM, XML
• Text: TXT, MD, JSON, YAML, YML
• Images: JPG, JPEG, PNG, TIFF, BMP, GIF (with OCR)
• Audio: MP3, WAV, FLAC, M4A, OGG (speech-to-text)
• Video: MP4, AVI, MOV, MKV, WEBM

Development

# Install in development mode
uv pip install -e .

# Run tests
python -m pytest tests/