Pdfnative

- Pure TypeScript, zero external dependencies.
- Generates 800+ pages in under 4 seconds.
- Supports ISO-compliant PDF/A and PKCS#7 digital signatures.
- 100% local processing — no external API calls.
- Tools to generate structured PDFs, add barcodes, and handle Unicode text.

- generate_basic_pdf: Generate a multi-page A4 PDF from structured blocks (headings, paragraphs, lists, page breaks, spacers). DEFAULT TOOL for plain documents — prefer this over specialized tools unless you need barcodes, tables, attachments, or non-Latin scripts. Optional pdfA flag enables Tagged PDF / PDF/A-1b/2b/2u/3b output (auto-embeds Noto Sans for non-WinAnsi Latin per ISO 19005 §6.3.4). Returns the PDF as base64 by default, or writes it to a sandboxed file path when outputMode=file.
- add_barcode: Generate a single-page PDF embedding a barcode or QR code. Supported formats:
• qr — URLs, vCards, any UTF-8 text ≤ 4296 chars. Use ecLevel='H' for printed media (logos/dirt-tolerant); 'M' (default) for screens.
• code128 — alphanumeric SKUs, ASCII payloads.
• ean13 — retail product codes (must be 12 or 13 digits; 13th is auto-computed).
• datamatrix— dense industrial / aerospace markings.
• pdf417 — ID cards, boarding passes.
Common recipe for a QR code pointing to a URL: { format: 'qr', data: 'https://example.com', caption: 'Scan me' }. The data field is the raw payload — do NOT pre-encode URLs.
- sign_pdf: Apply a PAdES-compatible CMS digital signature to a PDF. Since v1.0.0 you can sign ANY PDF in ONE call — autoInjectPlaceholder defaults to true, so you do NOT need to run prepare_signature_placeholder first unless you want to customize the placeholder appearance. Supports RSA-SHA256 and ECDSA-SHA256 (P-256). Required inputs: pdfBase64, algorithm, certDerBase64, plus EITHER rsaKeyPkcs1DerBase64 (when algorithm='rsa-sha256') OR ecPrivateScalarHex / ecPrivateKeyDerBase64 (when algorithm='ecdsa-sha256'). To convert PEM keys to DER base64: openssl pkey -in key.pem -outform DER | base64 -w0. To convert a PEM X.509 cert: openssl x509 -in cert.pem -outform DER | base64 -w0. After signing, call verify_pdf to confirm validity.
- add_international_text: Generate a PDF rendering text in any of 24 scripts (Arabic, Hebrew, Thai, CJK, Devanagari, Bengali, Tamil, Telugu, Sinhala, Tibetan, Khmer, Myanmar, Ethiopic, Cyrillic, Greek, Georgian, Armenian, Vietnamese, Turkish, Polish, Latin fallback) with optional COLRv1 colour emoji and mathematical / technical symbols (the "math" font — Noto Sans Math, ∀ ∃ √ ∑ ∫ ∞ ± ÷ ×). BiDi reordering (incl. UAX#9 isolates), Arabic harakat positioning, and complex-script OpenType shaping are handled automatically by the embedded Noto fonts; input is NFC-normalised for maximal glyph coverage and embedded newlines auto-split into paragraphs. Pass lang as a single code or an array (e.g. ["ar","emoji"] or ["latin","math"]) for multi-script / symbol runs.
- add_table: Generate a tabular PDF report from column headers and data rows. Ideal for data exports, financial summaries, schedules. Smart-table fields (pdfnative v1.2) automatically engage the document backend: wrap (auto/always/never), repeatHeader (header row on every page), zebra (alternate-row tint), caption (above the table, tagged for PDF/A), minRowHeight (points), cellPadding (points). Every row must have the same length as headers. For PDF/A output, set pdfA="pdfa2b" (most compatible).
- add_form: Generate a PDF containing an interactive AcroForm with text fields, text areas, checkboxes, radio buttons, and dropdowns. Suitable for data-capture forms, surveys, and fillable templates.
- embed_image: Generate a PDF document with an embedded JPEG or PNG image. The image is accepted as a base64-encoded string and can include an optional caption and custom render dimensions.
- prepare_signature_placeholder: Create a PDF with an embedded /Sig AcroForm placeholder ready to be digitally signed by the sign_pdf tool. NOTE: as of v1.0.0, sign_pdf auto-injects a placeholder when missing (autoInjectPlaceholder defaults to true), so this tool is OPTIONAL. Use it only when you need to: (a) customize the placeholder size for >4096-bit RSA keys via placeholderBytes, (b) attach the signature widget to a specific page via pageIndex, or (c) precompute and ship the placeholder PDF separately from the signing step. Otherwise call sign_pdf directly with any PDF.
- inspect_pdf: Read-only inspection of an existing PDF: version, page count, encryption state, PDF/A claim, signature count, hasSignaturePlaceholder, embedded attachments[], document info / metadata. Use the check array for CI-style assertions — supported values: 'pdfa', 'signed' (true when at least one signature has signed content), 'encrypted', 'placeholder' (unsigned /Sig widget present), 'attachments' (at least one /EmbeddedFile). The checksPassed boolean is true only when ALL requested checks hold.
- verify_pdf: Read-only verification of every PAdES Baseline / adbe.pkcs7.detached signature in a PDF. For each /Sig widget, recomputes the ByteRange SHA-256, validates the CMS messageDigest (integrity), and verifies the CMS signatureValue with the embedded signer certificate. Supports RSA-SHA256 and ECDSA-SHA256 (P-256). The response shape: { allValid, signatureCount, summary, signatures: [{ valid, integrity, signerSubject, signingTime, reason, chainTrust: 'self-signed'|'unverified'|'trusted', errors: [] }] }. Read allValid for an overall yes/no; iterate signatures[] for per-signature detail. Without trustedRootsDerBase64, chainTrust is 'self-signed' (single-cert chain) or 'unverified' (signer rooted in an external CA).
- add_attachment: Generate a PDF/A-3 (ISO 19005-3) document with one or more embedded files. USE THIS INSTEAD OF generate_basic_pdf when you need a Factur-X / ZUGFeRD electronic invoice (single XML payload with relationship='Source'), or any PDF that must carry machine-readable side-files. The visible document body is supplied via the optional blocks parameter (same block schema as generate_basic_pdf). The tool auto-emits PDF/A-3b conformance — PDF/A-3 is the only PDF/A part that legally permits embedded files. Each attachment is capped at 8 MiB.
- extract_text: Best-effort plain-text extraction from a non-encrypted PDF. Walks each page's content stream and pulls the operands of Tj/'/"/TJ text operators. The result.extractable boolean is FALSE when one or more pages have non-empty content but yielded no text (this is EXPECTED for PDFs using subset fonts without /ToUnicode CMaps — it is not an error). The accompanying extractableReason field explains why. Encrypted PDFs are rejected with EXTRACTION_UNSUPPORTED. Tagged-mode structure-tree extraction (cleaner output for tagged PDFs) is tracked on the roadmap.
- validate_pdf: Read-only PDF/UA (ISO 14289-1) structural conformance check. Verifies the accessibility prerequisites of a Tagged PDF: catalog /MarkInfo /Marked true, /StructTreeRoot (+ /ParentTree), /Metadata (XMP), /Lang, and per-page MCID uniqueness. Response shape: { standard: 'pdf-ua-1', valid, errors: [], warnings: [], summary }. Read valid for an overall yes/no; iterate errors[] for blocking violations and warnings[] for best-practice recommendations. This is a fast structural gate, NOT a full reference validator (veraPDF) — it does not check fonts, colour or rendering. Generate accessible input with any document tool using pdfA (e.g. pdfA='pdfa2u'), then validate the result here.
- extract_attachments: Read-only extraction of embedded files from a non-encrypted PDF (PDF/A-3 / Factur-X / ZUGFeRD). Walks the catalog /Names → /EmbeddedFiles tree and returns each attachment's metadata (name, mimeType, AFRelationship, description, sizeBytes) plus, by default, its decoded payload as dataBase64. Completes the invoice round-trip: add_attachment → inspect_pdf → extract_attachments. Pass filename to pull a single named file, or includeData: false for a metadata-only probe. Encrypted PDFs are rejected with EXTRACTION_UNSUPPORTED.
- merge_pdfs: Concatenate 2–50 source PDFs into a single document (pdfnative v1.4 page-tree API). Each kept page's object graph is deep-copied into a fresh, self-contained PDF. Signatures and AcroForms are dropped (a page-tree edit invalidates /ByteRange); self-contained URI link annotations are preserved unless dropAnnotations=true. Encrypted sources are rejected (ENCRYPTED_SOURCE) — decrypt first. A secure-by-default 256 MiB in-memory assembly guard (maxOutputSizeBytes) guards against memory exhaustion; the emitted PDF is separately capped at 50 MiB (OUTPUT_TOO_LARGE). Returns one PDF (base64 or sandboxed file).
- split_pdf: Split one PDF into several documents — one per requested page range (pdfnative v1.4 page-tree API). Ranges are 0-based and inclusive; end defaults to start (a single page). Each output is a fresh, self-contained PDF (signatures/AcroForm dropped; URI links kept unless dropAnnotations=true). Encrypted sources are rejected (ENCRYPTED_SOURCE). In base64 mode every produced PDF is returned as its own embedded resource block; in file mode each is written to a 1-based indexed sibling of outputPath ('out.pdf' → 'out-1.pdf', 'out-2.pdf', …). Use extract_pages instead when you want a single PDF from an arbitrary page subset.
- extract_pages: Extract an arbitrary subset of pages (0-based, in the order given) from a PDF into a SINGLE new document (pdfnative v1.4 page-tree API). The output is a fresh, self-contained PDF (signatures/AcroForm dropped; URI links kept unless dropAnnotations=true). Encrypted sources are rejected (ENCRYPTED_SOURCE). Use split_pdf instead when you need several output PDFs (one per range).
- annotate_pdf: Add markup / drawing annotations (ISO 32000-1 §12.5) to an existing PDF via pdfnative v1.5's annotation writer. Non-destructive incremental update: original content is preserved byte-for-byte and each annotation is appended to the target page's /Annots. Types: text (sticky note), highlight | underline | strikeout | squiggly (text-markup), square | circle (shapes), line, freetext. Each annotation needs a 0-based page and a rect [x1,y1,x2,y2]; line also needs start/end. Optional per-annotation: contents, color, opacity, title, plus type-specific fields (open/icon, quadPoints, interiorColor/borderWidth, fontSize). Encrypted sources are rejected (ENCRYPTED_SOURCE). NOTE: annotations are visual overlays — they do NOT remove or redact the underlying content.
- draft_governance_issue: Produce a LOCAL, governance-compliant GitHub issue draft plus a structured compliance report — and NEVER submit anything. This is the MCP-native embodiment of the pdfnative AI-governance / Human-In-The-Loop contract (.github/ai-governance.json, .github/AGENT_RULES.md): the agent is a DRAFTSMAN, the human is the only gate. The server makes NO outbound network call and has NO GitHub write path. The assembled draft is validated against the zero-dependency + reproduction policy; a violation (proposing a runtime dependency, missing reproduction, or duplicateSearchPerformed=false) throws GOVERNANCE_VIOLATION so the human must fix it before submitting under their own identity. Returns the draft markdown inline by default; outputMode='file' also writes a .md to the sandbox. After calling this, present BOTH the draft and the compliance report to the user, then STOP.
- read_form_fields: Read-only enumeration of an existing PDF's interactive AcroForm fields (pdfnative v1.6.0). Returns each terminal field's fully-qualified name, classified type (text | checkbox | radio | dropdown | listbox | button | signature | unknown), current value, flags (readOnly / required / multiline), choice options, and widget placements. Call this FIRST to discover field names before driving fill_form. Encrypted sources are supported via password. Token-frugal: verbosity:'summary' returns just { fieldCount }.
- fill_form: Fill (and optionally flatten) the AcroForm of an EXISTING PDF (pdfnative v1.6.0) — the counterpart to add_form, which CREATES a new form. Non-destructive incremental update: original bytes are preserved (a prior signature stays valid for its revision). values maps fully-qualified field name → value: text/choice take a string (array for multi-select listboxes); checkbox/radio take a boolean or the export-state string. Set flatten:true to stamp appearances into page content and drop the interactive layer (pass no values + flatten:true for a pure flatten). Encrypted documents are supported via password (appended objects are encrypted under the existing scheme). Signature fields cannot be filled (FORM_UNSUPPORTED). Discover field names with read_form_fields first.
- add_chart: Generate a single-page PDF with a native vector chart (pdfnative v1.6.0): bar, barH (horizontal bar), line (optional markers), pie or donut — rendered as pure PDF path operators, zero rasterisation. Multi-series bar/line, legends, 'nice' 1/2/5×10ⁿ axis ticks, gridlines, negative values, and a tagged-PDF /Figure + /Alt (auto-generated when altText omitted, so PDF/A stays conformant). Pie/donut use exactly one series (each value = a slice). Colours are hex strings (e.g. '#3366cc'). For a chart embedded amongst headings/paragraphs/tables, use a 'chart' block inside generate_basic_pdf instead.
- encrypt_pdf: Re-secure an existing PDF with the PDF Standard Security Handler (pdfnative v1.6.0): AES-128 (V4/R4, default) or AES-256 (V5/R6). RC4 is never emitted. Set ownerPassword (required) and optionally userPassword (open password), algorithm, and permissions { print, copy, modify, extractText }. Re-encrypt an already-encrypted source under a NEW password by passing its current password (password rotation in one call). CAVEAT: encryption rebuilds the page tree, so existing signatures and the interactive AcroForm are DROPPED and only self-contained URI links are kept — encrypt BEFORE signing, not after.
- decrypt_pdf: Open an encrypted PDF (pdfnative v1.6.0 reader/decryptor — RC4 V1–V4, AES-128 V4/R4, AES-256 V5/R6) and emit an UNENCRYPTED copy. Pass password (user or owner); documents with an empty user password decrypt without one. CAVEAT: decryption rebuilds the page tree, so existing signatures and the interactive AcroForm are DROPPED and only self-contained URI links are kept. To READ an encrypted PDF without rebuilding it, pass password directly to inspect_pdf / extract_text / extract_attachments instead.