PDF Tool

Read text from local PDF files, search inside PDFs, get metadata. Operates on filesystem only — no uploads.

Tools provided

pdf_read

Extract all text from a PDF.

{
  "tool": "pdf_read",
  "input": {
    "file_path": "/path/to/report.pdf",
    "max_pages": 50  // optional limit (0 = all)
  }
}

pdf_search

Find text within a PDF, returns surrounding context per match.

{
  "tool": "pdf_search",
  "input": {
    "file_path": "/path/to/report.pdf",
    "query": "valuation method",
    "context_chars": 100  // chars of context per match
  }
}

pdf_metadata

Page count, info dict, file size — without parsing full text.

{
  "tool": "pdf_metadata",
  "input": { "file_path": "/path/to/report.pdf" }
}

Output (pdf_read)

{
  "success": true,
  "data": {
    "file": "/abs/path/report.pdf",
    "pageCount": 42,
    "textLength": 84200,
    "info": { "Title": "...", "Author": "..." },
    "text": "Full extracted text..."
  }
}

Privacy

PDFs read from your local filesystem. The agent's filesystem sandbox rules still apply. Nothing uploaded.

Source

Implementation: packages/plugins/pdf/ (uses pdf-parse)