open-pdf

name	open-pdf
description	PDF parsing and extraction via MCP
version	1.0.0
metadata	{"tags":["pdf","parsing","extraction","document","mcp"],"category":"document","related_skills":["open-code"]}

Open PDF (PDF Parsing MCP Integration)

Extract text, tables, and metadata from PDF documents using the PDF parsing MCP server.

When to Use

• Extract text from PDF documents
• Parse tables and structured data
• Extract metadata (author, creation date, etc.)
• Convert PDF to text for analysis
• Process invoices, reports, contracts

When NOT to Use

• PDF generation (use browser PDF generation instead)
• Image extraction from PDFs (use specialized tools)
• PDF editing or modification
• Scanned PDFs without OCR capability

Prerequisites

• PDF MCP Docker container running (docker-compose.mesh.yml includes pdf-mcp)
• MCP server accessible at http://pdf-mcp:3200
• Agent profile with document_processing capability

Architecture

OpenOrchestrator (plan with PDF step)
  → OpenAgents (document-analyst profile)
  → MCP call to pdf-mcp-server
  → PDF container (parsing engine)
  → Results → OpenRec (audit) → OpenBrain (observation)

PDF MCP Tools

Basic Parsing

pdf_parse(file_path: string)

Extract all text content from PDF.

Returns:

{
  "text": "Full text content...",
  "pages": 10,
  "metadata": {
    "title": "Document Title",
    "author": "Author Name",
    "creationDate": "2024-01-01"
  }
}

Table Extraction

pdf_extract_tables(file_path: string, page?: number)

Extract tables from PDF (specific page or all pages).

Returns:

{
  "tables": [
    {
      "page": 1,
      "headers": ["Column 1", "Column 2"],
      "rows": [
        ["Value 1", "Value 2"],
        ["Value 3", "Value 4"]
      ]
    }
  ]
}

Metadata Extraction

pdf_get_metadata(file_path: string)

Extract PDF metadata only.

Returns:

{
  "title": "Document Title",
  "author": "Author Name",
  "subject": "Document Subject",
  "creator": "Creator Application",
  "producer": "PDF Producer",
  "creationDate": "2024-01-01T00:00:00Z",
  "modificationDate": "2024-01-02T00:00:00Z",
  "pageCount": 10
}

Page-Specific Extraction

pdf_extract_page(file_path: string, page_number: number)

Extract text from specific page.

Procedure

Basic Text Extraction

1. pdf_parse(file_path="/path/to/document.pdf")
2. Process extracted text
3. Store results in OpenBrain

Table Processing

1. pdf_extract_tables(file_path="/path/to/invoice.pdf")
2. Parse table structure
3. Convert to structured data (JSON/CSV)
4. Validate extracted data

Metadata Analysis

1. pdf_get_metadata(file_path="/path/to/document.pdf")
2. Extract document properties
3. Use for categorization or routing

Decision Rules

Scenario	Action
Need full text extraction	Use pdf_parse
Need structured table data	Use pdf_extract_tables
Need document properties	Use pdf_get_metadata
Need specific page only	Use pdf_extract_page
PDF is scanned/image-based	Flag for OCR tool (not supported)
PDF is encrypted/password-protected	Request password from user

Pitfalls

• Don't use for scanned PDFs without OCR
• Don't assume table extraction works for all formats
• Don't process very large PDFs (>100 pages) in single call
• Don't ignore metadata - it provides valuable context
• Don't forget to validate extracted data

Verification

• PDF MCP container running (docker ps | grep pdf-mcp)
• Can extract text from sample PDF
• Can extract tables from structured PDF
• Metadata extraction works
• Results logged to OpenRec

Security

• PDF parsing requires document_processing capability
• File paths must be within allowed directories
• No external network calls during parsing
• All operations logged to OpenRec

Integration with Mesh

// OpenOrchestrator plan step
{
  "goal": "Extract invoice data from PDF",
  "required_skills": ["open-pdf"],
  "approval_required": false
}

// OpenAgents dispatch
POST /v1/runs
{
  "profile": "document-analyst",
  "goal_id": "...",
  "parameters": {
    "file_path": "/data/invoices/INV-2024-001.pdf",
    "extract_tables": true
  }
}

// OpenRec audit
{
  "type": "pdf.parsing.completed",
  "payload": {
    "file": "INV-2024-001.pdf",
    "pages": 2,
    "tables_extracted": 1,
    "duration_ms": 450
  }
}

Related Skills

• open-code - Code execution for post-processing
• open-browser - PDF generation from web pages
• open-toolbox - Discover additional document tools