pdf-workflow

PDF Workflow

Orchestrates common multi-step PDF pipelines using the pdf-toolkit MCP server. Each workflow calls the 18 MCP tools in a specific order.

When to use

• Assemble: merge many PDFs then optimize the result
• Convert + secure: turn an Office doc into a watermarked, password-protected PDF
• Scan cleanup: repair a broken scan, OCR it, add page numbers
• Prep for email: aggressively compress + lock with a password
• Extract + OCR: pull specific pages then OCR them into Word

Assumed tools

• merge_pdf, split_pdf, compress_pdf
• pdf_to_word, pdf_to_excel, pdf_to_powerpoint, pdf_to_jpg
• office_to_pdf, html_to_pdf, image_to_pdf
• rotate_pdf, add_page_numbers, extract_pdf_pages
• unlock_pdf, protect_pdf, repair_pdf
• add_watermark, ocr_pdf

All tools accept input_file or input_files (absolute or CWD-relative paths) and an optional output_path.

Workflow 1: Assemble

Goal: combine N PDFs into one, then shrink, and optionally protect.

1. Call merge_pdf(input_files=[...], output_path="<tmp>/merged.pdf").
2. Call compress_pdf(input_file="<tmp>/merged.pdf", compression_level="recommended", output_path=<final>).
3. Optionally call protect_pdf(input_file=<final>, password=<user-provided>).

Workflow 2: Convert and secure

Goal: turn a Word/Excel/PPT file into a watermarked, password-protected PDF.

1. office_to_pdf(input_file=<file>) → <raw.pdf>.
2. add_watermark(input_file=<raw.pdf>, mode="text", text="CONFIDENTIAL") → <wm.pdf>.
3. protect_pdf(input_file=<wm.pdf>, password=<pw>) → final.

Workflow 3: Extract and OCR

Goal: pull specific pages from a PDF, OCR them, then convert to Word.

1. extract_pdf_pages(input_file=<src>, pages="3,5-9") → <sub.pdf>.
2. ocr_pdf(input_file=<sub.pdf>, languages=["eng"]) → <ocr.pdf>. (warn: slow for many pages)
3. pdf_to_word(input_file=<ocr.pdf>) → final.docx.

Workflow 4: Prep for email

Goal: shrink a PDF aggressively and lock it behind a password.

1. compress_pdf(input_file=<src>, compression_level="extreme") → <small.pdf>.
2. protect_pdf(input_file=<small.pdf>, password=<pw>) → final.

Workflow 5: Scan cleanup

Goal: fix a corrupt scan, OCR it, then number the pages.

1. repair_pdf(input_file=<scan>) → <fixed.pdf>.
2. ocr_pdf(input_file=<fixed.pdf>, languages=["eng"]) → <ocr.pdf>.
3. add_page_numbers(input_file=<ocr.pdf>, starting_number=1, horizontal_position="center", vertical_position="bottom") → final.

Notes

• If credentials are missing the tool call fails with MISSING_CREDENTIALS. Tell the user to set ILOVEAPI_PROJECT_PUBLIC_KEY and ILOVEAPI_PROJECT_SECRET_KEY env vars and restart Claude Code.
• ocr_pdf and large-PDF operations can exceed 60s. Mention that to the user before kicking off.
• Always prefer absolute paths. If the user gave relative paths, resolve against their CWD.
• Always report the final output path back to the user in your reply.