Actions

The connector exposes 33 actions. Every action takes a file input labelled by type — PDF File, Word File, Image File, and so on — plus, for some actions, a dropdown or a JSON options field. Actions return either a file or a JSON result.

​

File outputs

Actions that return a file expose the same three tokens in the dynamic content picker: Drop File Content into the next step — another Nitro action, Create file, or an email attachment. The Accept (Output format) input under Advanced parameters switches an action from returning the file (application/octet-stream, the default) to returning a short-lived download URL (application/json). See Output modes.

​

Conversions

Convert between PDF and other document formats. All return a file. Notes:

• PDF to Image is a multi-file output. In the default mode the result is a ZIP archive; in application/json mode it is a result.files array. See Multi-file outputs.
• PDF to PDFA conformance levels are 1b, 1a, 2b, 2u, 2a, 3b, 3u, 3a. The a levels require a tagged source PDF; use b or u for general archiving.

​

Transformations

Modify or restructure a PDF. All return a file. Notes:

• Compress has no options and always targets the smallest file. Use Optimize when you want to choose the size/quality trade-off.
• Flatten is irreversible — form fields, annotations, signatures, layers and transparencies become static page content.
• Rotate takes page_index (with an underscore), unlike every other action, which uses pageIndices. Valid amount values are 90, 180 and 270.
• Split takes an array of arrays: each inner array is the set of pages that becomes one output file. It is a multi-file output like PDF to Image.
• Merge adds a bookmark for each source document (Document 1, Document 2, …) by default, so readers can jump between them; no page is inserted. Pass {"tableOfContents":{"enabled":false}} to merge without bookmarks. To merge a variable number of files, see Working with files.
• Watermark boxes are [x0, y0, x1, y1] corner coordinates; Redact boxes are [x, y, width, height]. Both are in points from the top-left — see the Bounding Box Guide.
• Password Protect needs at least one of ownerPassword or userPassword. Permissions are print, modify, copy, annotate, form, assemble, print-hq.
• OCR processes one language per request: english, german, french, spanish, italian, finnish, swedish, danish, norwegian, dutch, portuguese, brazilian.

​

Extractions

Pull data out of a PDF. All of these return JSON under a result property, and none has an Accept input. Notes:

• Extract all Text with readingOrder: true reorders text to follow the visual reading flow. It helps on multi-column layouts but can interleave adjacent columns on dense designs.
• Extract searched Text matches literally and case-sensitively unless isRegex is true; regex queries accept regexFlags of ignore-case, multiline and dot-all.
• Extract PII and Extract searched Text return bounding boxes in the shape Redact PDF Pages accepts, so they chain into a detect-then-redact flow — see Feeding one action’s JSON into another.
• Extract Invoice Data handles scanned invoices with built-in OCR.
• Extract PDF Table Data and Smart Detect Form Fields have out-of-date token lists in the picker; read their results with an expression as described on Working with files.

​

Forms

Notes:

• Fill PDF Form accepts the values either as a file or inline in Options (JSON); provide one or the other. strict: true fails the request on unknown field names. Leave Accept at its default for this action — see Output modes.
• Create Fillable Forms takes exactly what Smart Detect Form Fields returns, so the two chain with a single expression.

​

Worked examples

These three pages document one action each in full — inputs, outputs, and a complete flow.