MCP PDF Forms

Instructions

Extract text from PDF pages

Args:
    pdf_path: Path to the PDF file
    start_page: Page number to start extraction (0-indexed). If None, starts from first page.
    end_page: Page number to end extraction (0-indexed, inclusive). If None, ends at start_page if specified, otherwise extracts all pages.
    
Returns:
    If extracting a single page: string containing the page text
    If extracting multiple pages: dictionary mapping page numbers to page text

Implementation Reference

• mcp_pdf_forms/server.py:247-303 (handler)

  The handler function for the 'extract_text' MCP tool. It uses PyMuPDF (fitz) to open the PDF and extract text from specified page ranges, returning either a string for single page or a dict of page texts for multiple pages. Registered via @mcp.tool() decorator.

  @mcp.tool()
  def extract_text(pdf_path: str, start_page: Optional[int] = None, end_page: Optional[int] = None) -> Union[str, Dict[int, str]]:
      """
      Extract text from PDF pages
      
      Args:
          pdf_path: Path to the PDF file
          start_page: Page number to start extraction (0-indexed). If None, starts from first page.
          end_page: Page number to end extraction (0-indexed, inclusive). If None, ends at start_page if specified, otherwise extracts all pages.
          
      Returns:
          If extracting a single page: string containing the page text
          If extracting multiple pages: dictionary mapping page numbers to page text
      """
      try:
          doc = fitz.open(pdf_path)
          total_pages = len(doc)
          
          # Validate page parameters
          if start_page is not None and (start_page < 0 or start_page >= total_pages):
              raise ValueError(f"Start page {start_page} is out of range (0-{total_pages-1})")
              
          if end_page is not None and (end_page < 0 or end_page >= total_pages):
              raise ValueError(f"End page {end_page} is out of range (0-{total_pages-1})")
              
          # Set defaults if parameters are None
          if start_page is None:
              start_page = 0
              
          if end_page is None:
              if start_page is not None:
                  end_page = start_page
              else:
                  end_page = total_pages - 1
                  
          # Ensure start_page <= end_page
          if start_page > end_page:
              start_page, end_page = end_page, start_page
          
          # Extract text
          if start_page == end_page:
              # Single page extraction
              page = doc[start_page]
              text = page.get_text()
              doc.close()
              return text
          else:
              # Multiple page extraction
              result = {}
              for page_num in range(start_page, end_page + 1):
                  page = doc[page_num]
                  result[page_num] = page.get_text()
              
              doc.close()
              return result
      except Exception as e:
          raise Exception(f"Error extracting text: {str(e)}")