pdf-writer-mcp
Creates, edits, and repairs PDFs via MCP, with accessible, archival, and signature-aware options.
Create PDFs from plain text, Markdown, or tabular data, with wrapping, pagination, margins, page sizes, fonts, CJK font embedding, and missing-glyph handling.
Write PDF 1.7 or PDF 2.0 output, optionally tagged for PDF/UA accessibility, with deterministic output via SOURCE_DATE_EPOCH.
Return a saved file path or base64 data.
Edit metadata, bookmarks, annotations, page numbers, watermarks, and embedded attachments.
Merge, split, extract, delete, reorder, and rotate pages.
Fill or flatten AcroForm fields; repair form tagging for PDF/UA.
Scaffold untagged PDFs into PDF/UA-1 or add PDF/A-3b, PDF/A-4, or PDF/A-4f declarations, with warnings that conformance still needs verification.
Preserve digital signatures through incremental updates where supported, or explicitly allow breaking them.
Enforce absolute paths, report structured errors with next actions, and return warnings when information is lost or assumptions are made.
Converts Markdown documents into PDFs via create_markdown_pdf, supporting headings, paragraphs, bullet/ordered lists, code blocks, quotes, horizontal rules and tables. When tagging is enabled, Markdown constructs map directly onto the PDF/UA structure tree (headings → H1–H6, lists → L/LI/LBody, tables → Table/TR/TH/TD, quotes → BlockQuote, code → Code), with heading levels normalised so they start at H1 and never skip.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@pdf-writer-mcp Create a tagged PDF from /tmp/report.md with title 'Q3 Report'"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
pdf-writer-mcp
MCP server that creates PDFs from text, Markdown, or tabular data and edits existing ones (metadata and page operations). Built on normativepdf — a clause-driven PDF library in which every behaviour is tied to an ISO 32000 clause — with CJK font embedding via harfbuzz subsetting.
Part of the PDF family alongside pdf-reader-mcp (structure analysis) and pdf-verify-mcp (authenticity verification). Where pdf-reader-mcp tells you what is in a PDF and pdf-verify-mcp tells you whether it is genuine, pdf-writer-mcp is the one that writes it.
Tools
All file paths must be absolute (since v0.7.0). Relative paths and paths containing .. are rejected — a relative path would resolve against the MCP host's working directory, which is not the directory you think it is. This applies to inputPath, inputPaths, outputPath, outputDir, fontPath and attachmentPath. Input PDFs larger than 100 MB are also rejected.
Creation
Tool | Purpose |
| Plain text — honours |
| Markdown — headings, paragraphs, bullet/ordered lists, code blocks, quotes, rules, tables |
| Ruled tables — automatic column widths, cell wrapping, headers repeated across page breaks |
Shared options: outputPath, returnBase64, fontPath, fontSize, pageSize (A4/A3/A5/LETTER/LEGAL), margin, title, drawTitle, author, onMissingGlyph, tagged, lang, pdfVersion.
PDF 2.0 output (v0.16.0)
Pass pdfVersion: "2.0" to write an ISO 32000-2 file. The default stays "1.7", and its bytes are unchanged.
The version is not only a header. ISO 32000-2 attaches two obligations to it, and both are met:
trailer
/IDbecomes Required (Table 15). Both elements are equal on a first write (R-14.4-6), and the value stays deterministic underSOURCE_DATE_EPOCH.the Info dictionary keeps only
CreationDateandModDate(§14.3.3); the title, author and producer move to the XMP metadata stream, where PDF 2.0 says document metadata belongs.
{ "text": "Body.", "title": "Report", "pdfVersion": "2.0" }With tagged: true, a 2.0 file declares PDF/UA-2 (ISO 14289-2) instead of PDF/UA-1, which allows only PDF 1.0 to 1.7 (ISO 14289-1 §6.1). See PDF/UA-2 below. Up to v0.26.x the combination was refused.
Tagged PDF / PDF/UA (v0.5.0)
Pass tagged: true to produce an accessible, tagged PDF conforming to PDF/UA-1 (ISO 14289). Output is verified compliant by veraPDF (--flavour ua1, 106/106 rules).
{ "markdown": "# Title\n\nBody.", "title": "Report", "tagged": true, "lang": "en" }Markdown maps onto the structure tree: headings → H1–H6, lists → L/LI/LBody, tables → Table/TR/TH/TD (headers get /Scope), quotes → BlockQuote, code → Code. Rules, borders and code backgrounds become artifacts. Under PDF/UA-1, heading levels are normalised so they start at H1 and never skip — a Markdown # → ### jump becomes H1 → H2 in the structure, while visual sizes stay as authored (PDF/UA-2 keeps the level as written; see below).
PDF/UA mandates a document title, so tagged: true requires title. The title is always written to the metadata (dc:title) and, by default, also drawn as the first H1. When the Markdown already starts with its own # … heading, pass drawTitle: false: the title then stays in the metadata only, and the first body heading becomes H1 (## → ### becomes H1 → H2). Output stays PDF/UA-1 compliant (veraPDF 106/106). lang (BCP 47) is inferred from the text when omitted and reported via warnings — pass it explicitly when you know it, since a wrong /Lang makes screen readers mispronounce the text.
Tagging is opt-in — default output is unchanged.
Machine validation (veraPDF) only sees whether things exist. It cannot judge whether reading order or alt text are appropriate — human review still matters.
PDF/UA-2 (v0.27.0)
Add pdfVersion: "2.0" to tagged: true and the file is written as PDF/UA-2 (ISO 14289-2:2024) on PDF 2.0. The default (1.7, PDF/UA-1) is unchanged.
{ "markdown": "# Title\n\nBody.", "title": "Report", "tagged": true, "pdfVersion": "2.0", "lang": "en" }What differs from the PDF/UA-1 output, with the clause behind each:
PDF/UA-1 (1.7) | PDF/UA-2 (2.0) | |
XMP declaration (§5) |
|
|
Namespaces (§8.2.4, §8.2.5.2) | none (the default PDF 1.7 namespace) |
|
Drawn title (§8.2.5.13) | first |
|
Markdown headings (§8.2.5.12) | normalised to start at H1 and never skip | the level as written ( |
Nested lists | inside the parent item's | directly in the parent |
Measured with veraPDF 1.30.0 (pdfua-2, 1,728 rules): the output of all three create tools, ensure_tagged on an untagged PDF 2.0 file, and ensure_pdfa(flavour: "pdfa-4") on a PDF/UA-2 file all pass, and the last one also passes PDF/A-4 (109/109).
Edits keep the declaration: when a tool rebuilds the XMP, pdfuaid:rev is carried with pdfuaid:part, and dc:title is kept when the Info dictionary has no Title (PDF 2.0 output does not put one there). add_bookmarks writes structure destinations on a PDF/UA-2 file, as §8.8 requires: a GoTo action whose SD names the structure element that starts on the target page, with D kept as the page destination for viewers that do not read SD. A page with no structure content keeps a page destination and is named in warnings.
Versions and declarations
An edit writes the PDF version of its input. Only an explicit conversion changes it: ensure_pdfa with flavour: "pdfa-4" / "pdfa-4f" (to PDF 2.0), and merge_pdfs with pdfVersion. Versions are never lowered.
Anything that would break a shall of the output's version, or of a standard the file declares, is refused with
INVALID_ARGUMENT. The message names the input's version and declarations;next_actionsgives a call that fits. Examples:merge_pdfsover PDF 2.0 and PDF 1.x inputs withoutpdfVersion, andensure_pdfa(flavour: "pdfa-3b")on any PDF 2.0 input (PDF/A-3 needs a%PDF-1.nheader).A form that the output's version deprecates is written the way that version prescribes. On a PDF 2.0 input,
set_metadatawrites to XMP and adds no Info entries (ISO 32000-2 §14.3.3); on a PDF/UA-2 input it also moves the existing Info entries to XMP and removes them (ISO 14289-2 §6.2).A mix the standard allows is written and reported. Structure elements that
add_annotationandtag_form_fieldsadd to a PDF/UA-2 file have no/NS, so they sit in the PDF 1.7 namespace (§8.2.4 allows it);warningssays so.
Every tool result carries output, read back from the written file:
{ "output": { "pdfVersion": "2.0", "declares": { "pdfua": 2, "pdfa": "4" } } }declares.pdfua is pdfuaid:part; declares.pdfa joins pdfaid:part and pdfaid:conformance ("3b", "4", "4f"). A key is absent when nothing is declared. A declaration is only a claim — measure it with pdf-verify-mcp.
When a tool rewrites the XMP (set_metadata, ensure_pdfa), the values you did not pass are carried over as written: the other languages of dc:title, every name in dc:creator, and properties in namespaces this server does not know. Changing title replaces the x-default entry and names the other languages in warnings.
Editing — page operations
Tool | Purpose |
| Concatenate 2–50 PDFs in order; metadata inherited from the first file. Inputs that mix PDF 2.0 and PDF 1.x need |
| One output file per page range |
| Extract pages in the requested order (doubles as reordering) |
| Remove pages (deleting every page is rejected) |
| Reorder by an explicit permutation of all pages |
| Rotate clockwise (90/180/270), accumulating over existing rotation. Edits in place, so the warning below does not apply |
The five tools other than rotate_pages rebuild the document from its pages.
Carried over: attachments (
/Names /EmbeddedFiles,/AF),/Lang,/ViewerPreferences,/OutputIntentsNot carried over: the tagged structure tree, XMP, and anything tied to page numbers or page references (bookmarks, page labels, named destinations)
Whatever is lost is reported in
warnings— nothing disappears silently
Page specs use "1,3-5,8-" (1-based; -3 means up to page 3, 8- means page 8 to the end). Order is preserved and duplicates are removed.
Editing — adding to and repairing documents
Tool | Purpose |
| Update Info dictionary fields ( |
| Set the outline (bookmarks); nestable via |
| Add a sticky note ( |
| Embed a file ( |
| Stamp page numbers ( |
| Fill AcroForm fields. Japanese values via an embedded font; can flatten in the same pass |
| Flatten a form into static content. Refuses tagged PDFs by default (breaks PDF/UA) |
| Repair the form inside a tagged PDF for PDF/UA-1 (nest widgets in |
| Put an existing PDF into the PDF/UA-1 container → Scaffolding an untagged PDF |
| Put an existing PDF into the PDF/A-3b / PDF/A-4 / PDF/A-4f container ( |
| Overlay a diagonal watermark ("社外秘" / "DRAFT"). Behind the body content by default; artifact on tagged PDFs |
Shared options: outputPath, returnBase64, allowBreakingSignatures.
Editing signed PDFs: an ordinary save rewrites the whole file, so editing normally invalidates existing signatures. PDFs containing /ByteRange are rejected by default.
preserveSignatures: true— appends an ISO 32000 incremental update that keeps every signature valid (the original bytes are untouched). Supported by every editing tool that adds to a document:add_annotation,set_metadata,add_bookmarks,tag_form_fields,ensure_tagged,attach_file,stamp_page_numbers,add_watermark(on tagged PDFs the structure-tree changes ride the same increment)allowBreakingSignatures: true— proceed destructively, invalidating signaturesCertified documents (DocMDP) are refused when the change type is not permitted by the certification level (§12.8.2.2)
Measured: stacked increments on a really-signed PDF keep pdf-verify-mcp reporting VALID, and incremental structure updates on tagged PDFs stay veraPDF COMPLIANT (106/106).
Scaffolding an untagged PDF (ensure_tagged)
ensure_tagged puts an existing PDF into the PDF/UA-1 container. On tagged input the structure tree is left untouched and only missing document-level requirements are repaired (MarkInfo, /Lang, DisplayDocTitle, XMP). On a document that never had a structure tree, a minimal scaffold is created — each page's content wrapped in a single P element, which makes the text reachable by assistive technology and passes veraPDF (measured: 106/106).
This is a scaffold, not accessibility. A machine cannot infer meaning, so headings, lists, tables, reading order and figure alt text are not produced. The tool says so in its warnings. (Wrapping the content in Artifact would also pass veraPDF while hiding the body from screen readers — conformance theatre, deliberately not implemented.) Where you control the source, create_* with tagged: true produces real structure; ensure_tagged is for documents you were handed.
The archival container (ensure_pdfa)
ensure_pdfa is the archival (PDF/A-3b) counterpart of ensure_tagged. It adds only the document-level requirements:
trailer
/ID(ISO 32000-1 §14.4)an sRGB output intent (
GTS_PDFA1; ICC profile generated and embedded)the XMP
pdfaiddeclaration (the creation date is inherited from Info/CreationDate)
Content streams, fonts and the structure tree are untouched — so thisdoes not make a PDF conform. Write the declaration, then measure it: verify with pdf-verify-mcp's validate_conformance(flavour: "pdfa-3b"). Measured on the electronic-bookkeeping sample: veraPDF 146/146 COMPLIANT, PDF/UA-1 still 106/106, attachment preserved.
Since v0.17.0 the result also carries declarationRisks when the claim is already known to be false. The only risk reported today is FONT_NOT_EMBEDDED: PDF/A requires every font to be embedded and ensure_pdfa does not embed fonts, so a document drawn with the standard 14 faces fails validation regardless of how correct its document-level structure is. The declaration is still written — the verdict belongs to veraPDF — but the caller can now branch on the risk instead of parsing the warning text.
Since v0.24.0 the result carries previousDeclaration — the PDF/A the input already claimed, as { part, conformance }, or null — in place of wasDeclared (a boolean that did not say whether the earlier claim matched the requested flavour). This is a breaking change.
ensure_pdfa refuses pdfa-4 / pdfa-4f for an input that declares PDF/UA-1 (v0.24.0). PDF/A-4 sets the header to PDF 2.0, while PDF/UA-1 allows versions 1.0 to 1.7 (ISO 14289-1 §6.1), so the file would fail one of its two claims. The error is INVALID_ARGUMENT with next_actions pointing at flavour: "pdfa-3b", which keeps the 1.7 header.
Since v0.27.0 a PDF/UA-2 input is the other way round: pdfa-4 / pdfa-4f are accepted (both standards are built on PDF 2.0), and pdfa-3b is refused with next_actions pointing at flavour: "pdfa-4" — PDF/A-3 requires a PDF 1.x header (veraPDF rule ISO 19005-3:2012 6.1.2-1) and PDF/UA-2 requires 2.0.
PDF/A-4 (v0.16.0)
flavour: "pdfa-4" targets ISO 19005-4 instead. It is built on PDF 2.0, so on top of the three items above it rewrites the header to 2.0 and removes the Info dictionary — PDF/A-4 does not allow one unless the catalog has /PieceInfo, which is stricter than ISO 32000-2 §14.3.3. pdfaid:rev is written and pdfaid:conformance is not, because PDF/A-4 has no conformance level.
If the document carries attachments, use "pdfa-4f". Plain PDF/A-4 requires every embedded file to be PDF/A itself, so attaching a CSV or JSON — the electronic-bookkeeping case — makes it non-conformant. "pdfa-4f" is the variant for exactly that.
Measured with veraPDF 1.30.0: pdfa-4 109/109 COMPLIANT, and the same document with a CSV attached is 108/109 as pdfa-4 but 109/109 as pdfa-4f.
preserveSignatures is refused with a PDF/A-4 flavour unless the input is already PDF 2.0: an incremental update cannot rewrite byte 0 of the file, and doing so anyway would break the signature it was meant to preserve.
Related MCP server: MCP MD2PDF Server
Install
{
"mcpServers": {
"pdf-writer": {
"command": "npx",
"args": ["-y", "@shuji-bonji/pdf-writer-mcp@latest"],
"env": {
"PDF_WRITER_FONT": "/absolute/path/to/NotoSansJP-Regular.otf"
}
}
}
}PDF_WRITER_FONT lets every tool omit fontPath and still render CJK text.
Use @latest (or pin a version). npx -y <pkg> without a version keeps running whatever it cached the first time — -y only skips the install prompt, it does not check for updates. A bare specifier will happily run a months-old release. @latest makes npx check the registry on each start; pin @0.5.0 instead if you want reproducibility. To clear a stale cache: rm -rf ~/.npm/_npx.
Fonts
The standard PDF font (Helvetica) covers the 218 characters of WinAnsiEncoding (ISO 32000-1 Annex D.2): ASCII, Latin-1 letters and signs, and curly quotes, dashes, €, …, ™ and a few more. Text with any other character from U+0100 up (Japanese, ł, ✔, …) is refused with FONT_REQUIRED, and the error names the characters. To render such text, point fontPath or PDF_WRITER_FONT at an embeddable single-face font (.ttf / .otf).
Recommended source: Noto Sans JP (SubsetOTF/JP) — static, single-face, SIL OFL.
.ttc(TrueType Collection) is not supported — the file is detected and rejected. Extract a single face first:python3 -c "from fontTools.ttLib import TTCollection; \ TTCollection('NotoSansCJK-Regular.ttc').fonts[0].save('NotoSansCJKjp-Regular.otf')"
Missing glyphs
Characters absent from the font (e.g. ✔ U+2714, which Noto Sans JP does not include) would otherwise be embedded as .notdef and render as silent blanks. onMissingGlyph controls this:
Value | Behaviour |
| Fail, listing the offending characters as |
| Substitute 〓 (or |
| Render as blanks and report via |
Result
{
"path": "/abs/out.pdf", // when outputPath is given
"base64": "JVBERi0xLj...", // when returnBase64, or outputPath is omitted
"pageCount": 3,
"bytes": 91788,
"font": "NotoSansJP-Regular.otf",
"warnings": ["Replaced 1 unsupported character(s) with \"〓\": \"✔\" (U+2714)"]
}Editing tools return the same shape without font; split_pdf returns { files: [...], count }.
Errors (v0.7.0)
Errors are structured, following the same contract as pdf-reader-mcp: a stable code for programs, plus next_actions an LLM agent can act on. Writer-specific guards are all expressed as retryable with an explicit flag:
{
"error": "\"/in/signed.pdf\" appears to be digitally signed (/ByteRange found). …",
"code": "SIGNED_PDF",
"retryable": true,
"next_actions": [
{
"action": "retry_with_allowBreakingSignatures",
"reason": "Only if invalidating the signature is acceptable…",
"example": { "allowBreakingSignatures": true }
}
]
}Codes: INVALID_ARGUMENT, DOC_NOT_FOUND, FONT_NOT_FOUND, INVALID_PDF, ENCRYPTED_PDF, UNSUPPORTED_PDF_FEATURE (XFA), FILE_TOO_LARGE, INTERNAL_ERROR, and the writer guards SIGNED_PDF (allowBreakingSignatures), TAGGED_PDF (allowBreakingTags), FONT_REQUIRED (fontPath), MISSING_GLYPH (onMissingGlyph).
Since v0.23.3 every error carries retryable, fixed per code (PDF family contract, pdf-agent-stack docs/specs/06 §2.3): true for INVALID_ARGUMENT, FONT_REQUIRED, FONT_NOT_FOUND, MISSING_GLYPH, SIGNED_PDF and TAGGED_PDF — changing an argument or a flag can make the same edit succeed — and false for the rest. error, hint and next_actions[].reason are English; the values of next_actions[].action and example do not change.
Deterministic output (v0.7.0)
Set the SOURCE_DATE_EPOCH environment variable (UNIX seconds, per the reproducible-builds.org convention) to pin CreationDate, ModificationDate and XMP timestamps. The same input then yields byte-identical output — useful for diffing, caching, and reproducible tests. Invalid values raise an error rather than being ignored.
Text extraction
Generated PDFs are selectable, copyable, searchable, and screen-reader accessible: a ToUnicode CMap (§9.10.3) is written even for embedded subset fonts. This is covered by regression tests (extract.test.ts, render.test.ts).
Output from v0.13.x and earlier could make poppler-based viewers printMismatch between font type and embedded font file. That was a symptom of a real conformance defect (W-2: CFF fonts embedded via FontFile2), fixed in v0.14.0 — current output produces no such warning.
Development
npm install
npm run build # emits dist/
npm test # vitest
npm run typecheck # tsc --noEmitFont-dependent tests activate when TEST_FONT_PATH points at a CJK font:
TEST_FONT_PATH=/path/to/NotoSansJP-Regular.otf npm testKnown limitations
Inline styling: bold/italic affect size and glyph text only, not typeface — a single font is embedded per document.
.ttcfonts require extracting a single face (see above).By default,
titleand the first body heading both become H1 (B-19). This satisfies PDF/UA 7.4.2 (start at H1, skip no levels) and passes veraPDF, but the heading is duplicated when the body repeats the title. PassdrawTitle: falseto keep the title in the metadata only.List
/Lblis not emitted (B-18). Bullets and numbers are baked into the body text. ISO 32000-2 §14.8.4.8.2 makesLbla NOTE (often include), not a requirement, so this conforms — but regenerating from a read-back duplicates the markers.
Resolved:
The subset name prefix (
ABCDEF+) is applied as of v0.14.0 (W-3). The earlier limitation stating otherwise no longer holds.Underscores in
snake_casebeing dropped (B-17) was fixed on 2026-07-21 (ships in the next release)._emphasis now requires a non-intraword position and code spans are protected first, so non-ASCII identifiers such as日本語_変数名survive too.
Roadmap
Editing Tier A wave 1 — metadata and page operations (v0.2.0)
Editing Tier A wave 2 — bookmarks and annotations (v0.4.0)
Tagged PDF / PDF/UA-1 — verified by veraPDF (v0.5.0)
Annotations nested in
Annottags on tagged output (v0.5.1)Editing Tier B — file attachments, form filling/flattening, watermarks, page-number stamping (v0.6.0)
Code hygiene / family alignment — McpServer + Zod, structured errors, absolute-path enforcement, stdout guard, tool annotations, deterministic output (v0.7.0)
tag_form_fields— PDF/UA repair for forms in tagged PDFs, verified COMPLIANT by veraPDF (v0.8.0)Tier C first milestone — signature-preserving incremental updates for
add_annotation, verified by pdf-verify-mcp against a real CMS signature (v0.9.0)Incremental updates extended to
set_metadata/add_bookmarks, full trailer carry-over (§7.5.6), XMP kept in sync with Info (v0.10.0)Incremental updates on tagged PDFs — generalised dirty tracking over the structure tree;
tag_form_fieldsgainspreserveSignatures(v0.11.0)Incremental updates across every editing tool, and
ensure_tagged— PDF/UA scaffold & repair (v0.12.0)Page operations report and carry over document-level information; three shall violations found by re-auditing against ISO 32000-2 (v0.13.0)
Carrying the structure tree through page operations (which also unlocks MarkInfo, conformance-declaring XMP,
/AcroFormand merging attachments across inputs)Tier C remainder —
edit_text(body text editing/reflow)Publish-pipeline skill (write → read back with pdf-reader → gate with pdf-verify)
Images with alt text (
Figure+/Alt)Automatic face extraction from
.ttcSeparate faces for headings and body (bold face embedding)
Image embedding, headers/footers
Tier C — signature-preserving incremental updates, body text editing, tag tree maintenance
PDF/A conversion
License
MIT © shuji-bonji
Available Tools
20 toolsadd_annotationAdd AnnotationA
Add one annotation to a page: sticky note (text), highlight, or rectangle (square). Coordinates are in PDF space (origin bottom-left, pt). For signed PDFs, preserveSignatures: true appends an incremental update without invalidating existing signatures (in tagged documents the enclosure in an Annot structure element rides the same update, preserving PDF/UA conformance).
Inputs: page (required), rect (required), type (required), inputPath (required), +11 more.
add_bookmarksAdd Bookmarks (Outline)AIdempotent
Set the bookmarks (outline) of a PDF. Existing bookmarks are replaced. Nest with children. For signed PDFs, preserveSignatures: true sets them while keeping the signatures intact.
Inputs: bookmarks (required), inputPath (required), outputPath, returnBase64, +2 more.
add_watermarkAdd WatermarkAIdempotent
Overlay a diagonal watermark across the middle of each page ("社外秘" / "DRAFT" / "COPY", etc.). Drawn faintly behind the content by default. In tagged PDFs it is wrapped as an Artifact, preserving PDF/UA conformance. CJK watermarks need fontPath or the PDF_WRITER_FONT environment variable.
Inputs: text (required), inputPath (required), angle, color, +9 more.
attach_fileAttach File (Embedded File)AIdempotent
Embed (attach) a file into a PDF. Registers it under /Names /EmbeddedFiles and the catalog /AF, with an AFRelationship. For PDF/A-3 (ISO 19005-3) and Japanese e-bookkeeping-law (電子帳簿保存法) workflows that bundle "a human-readable invoice PDF + machine-readable data (CSV/XML)" into one file.
Inputs: inputPath (required), attachmentPath (required), name, mimeType, +6 more.
create_markdown_pdfCreate PDF from MarkdownAIdempotent
Create a PDF from Markdown. Supports headings, paragraphs, bullet/numbered lists, code blocks, quotes, horizontal rules and tables. Inline decoration markers are stripped and the text rendered plain (single font).
Inputs: markdown (required), lang, title, author, +9 more.
create_table_pdfCreate Table PDFAIdempotent
Create a ruled table PDF from headers and row data. Column widths are computed from the content, cells wrap, and the header row is redrawn after page breaks.
Inputs: rows (required), headers (required), lang, title, +10 more.
create_text_pdfCreate PDF from Plain TextAIdempotent
Create a PDF from plain text. Honours line breaks (\n) and treats blank lines as paragraph breaks. Long lines wrap automatically.
Inputs: text (required), lang, title, author, +9 more.
delete_pagesDelete PagesADestructiveIdempotent
Create a new PDF with the given pages removed. Deleting every page is an error. Pages are copied into a new document, so document-level information (tagged structure, XMP, attachments, AcroForm, bookmarks, etc.) is not carried over. Anything lost is reported in warnings; follow up on the output with attach_file / ensure_tagged / add_bookmarks / set_metadata as needed.
Inputs: pages (required), inputPath (required), outputPath, returnBase64, +1 more.
ensure_pdfaEnsure PDF/A (archival conformance scaffold)AIdempotent
Put an existing PDF onto the PDF/A "vessel" (the PDF/A counterpart of ensure_tagged). Choose the flavour: "pdfa-3b" (default) / "pdfa-4" / "pdfa-4f". Supplies only the missing document-level requirements: the trailer /ID (ISO 32000-1 14.4), an sRGB OutputIntent (GTS_PDFA1; an ICC profile is generated and embedded), and XMP pdfaid. The -4 flavours additionally set the header to PDF 2.0 and delete the Info dictionary (-4 forbids Info unless the catalog has /PieceInfo — stricter than ISO 32000-2 14.3.3). Content, structure tree and fonts are never touched. Documents with attachments must use "pdfa-4f" — plain "pdfa-4" requires every attachment to be PDF/A itself, so bundling CSV or JSON (the Japanese e-bookkeeping-law pattern) would not conform. IMPORTANT: this is preparation for claiming PDF/A, not a guarantee of conformance. Violations such as unembedded fonts, encryption, JavaScript or LZW are not repaired. Writing pdfaid into XMP is the document claiming "I am PDF/A" — applied to a non-conforming document it produces a PDF that lies about itself (which is why a warning is always returned). Always confirm with pdf-verify-mcp's validate_conformance (flavour: the same value) — the verdict is veraPDF's, and since ISO 19005 clauses cannot be quoted, the strongest statement is "veraPDF judged it so". In the e-bookkeeping-law context, apply it after attaching machine-readable data with attach_file. For signed PDFs, preserveSignatures: true (approval signatures only; certification signatures are refused). However, the -4 flavours combined with preserveSignatures are refused unless the input is already PDF 2.0 (an incremental update cannot rewrite the file header, and rewriting it would break the signatures).
Inputs: inputPath (required), flavour, outputPath, returnBase64, +2 more.
ensure_taggedEnsure Tagged (PDF/UA scaffold & repair)AIdempotent
Put an existing PDF onto the PDF/UA-1 "vessel". If it is already tagged, the structure tree is untouched and only missing document-level requirements are supplied (MarkInfo / Lang / DisplayDocTitle / XMP pdfuaid:part and dc:title). For untagged documents, a minimal structure tree (each page = one P element) is created so the content becomes reachable by assistive technology. IMPORTANT: machines cannot infer meaning — headings, tables, lists, reading order and figure alt text are NOT created. The new tree is a scaffold, not an accessible document; it needs human review. If you can build the structure right from the start, use the create tools' tagged: true. For signed PDFs, preserveSignatures: true (approval signatures only; certification signatures are refused).
Inputs: inputPath (required), lang, title, outputPath, +3 more.
extract_pagesExtract PagesAIdempotent
Create a new PDF containing only the given pages. The given order is preserved, so extraction doubles as reordering. Pages are copied into a new document, so document-level information (tagged structure, XMP, attachments, AcroForm, bookmarks, etc.) is not carried over. Anything lost is reported in warnings; follow up on the output with attach_file / ensure_tagged / add_bookmarks / set_metadata as needed.
Inputs: pages (required), inputPath (required), outputPath, returnBase64, +1 more.
fill_formFill Form (AcroForm)AIdempotent
Fill field values into an existing PDF's interactive form (AcroForm). If you do not know the field names, pass a nonexistent one — the error lists every field name and type. CJK values need fontPath or the PDF_WRITER_FONT environment variable. flatten: true makes the form non-interactive after filling, but on a tagged PDF that breaks PDF/UA conformance and additionally requires allowBreakingTags: true. XFA forms are not supported.
Inputs: fields (required), inputPath (required), flatten, fontPath, +4 more.
flatten_formFlatten FormADestructiveIdempotent
Flatten an existing PDF's interactive form (AcroForm), keeping the filled appearance while removing interactivity. Use it to freeze values before distribution. If existing values contain CJK text, set fontPath or PDF_WRITER_FONT in case appearances must be regenerated. On tagged PDFs, Widget annotations disappear and Form structure elements are left dangling, so it refuses by default (allowBreakingTags: true to force).
Inputs: inputPath (required), fontPath, outputPath, returnBase64, +2 more.
merge_pdfsMerge PDFsAIdempotent
Merge multiple PDFs into one, in the given order. Document metadata is carried over from the first file. Pages are copied into a new document, so document-level information (tagged structure, XMP, attachments, AcroForm, bookmarks, etc.) is not carried over. Anything lost is reported in warnings; follow up on the output with attach_file / ensure_tagged / add_bookmarks / set_metadata as needed.
Inputs: inputPaths (required), outputPath, returnBase64, allowBreakingSignatures.
reorder_pagesReorder PagesAIdempotent
Reorder pages. order must list every page exactly once, in the new order. Pages are copied into a new document, so document-level information (tagged structure, XMP, attachments, AcroForm, bookmarks, etc.) is not carried over. Anything lost is reported in warnings; follow up on the output with attach_file / ensure_tagged / add_bookmarks / set_metadata as needed.
Inputs: order (required), inputPath (required), outputPath, returnBase64, +1 more.
rotate_pagesRotate PagesBIdempotent
Rotate pages clockwise (90/180/270 degrees). All pages when pages is omitted.
Inputs: rotation (required), inputPath (required), pages, outputPath, +2 more.
set_metadataSet PDF MetadataAIdempotent
Update an existing PDF's metadata (the Info dictionary). Only the given fields change; the rest are preserved. At least one of title / author / subject / keywords / creator is required. In documents with XMP (/Metadata), dc:title etc. are synchronized to prevent divergence. For signed PDFs, preserveSignatures: true updates while keeping the signatures intact.
Inputs: inputPath (required), title, author, creator, +6 more.
split_pdfSplit PDFAIdempotent
Split a PDF into multiple files by page range. Each element of ranges becomes one file, named "1.pdf", "2.pdf", and so on. Pages are copied into a new document, so document-level information (tagged structure, XMP, attachments, AcroForm, bookmarks, etc.) is not carried over. Anything lost is reported in warnings; follow up on the output with attach_file / ensure_tagged / add_bookmarks / set_metadata as needed.
Inputs: ranges (required), inputPath (required), outputDir (required), prefix, +1 more.
stamp_page_numbersStamp Page NumbersAIdempotent
Stamp a page number on each page. In tagged PDFs the stamp is wrapped as an Artifact, preserving PDF/UA conformance. Formats containing CJK text need fontPath or the PDF_WRITER_FONT environment variable.
Inputs: inputPath (required), color, pages, format, +9 more.
tag_form_fieldsTag Form Fields (PDF/UA repair)AIdempotent
Repair a tagged PDF's form to PDF/UA-1: enclose Widget annotations in Form structure elements (7.18.4-1), set /Tabs S on the affected pages (7.18.3-1), and give fields alternate names /TU (7.18.1-3). Pass human-readable names for screen readers via labels. Widgets already bound to the structure tree are skipped, so it is safe to run repeatedly. Untagged documents are out of scope (rebuild with the create tools' tagged: true, or run ensure_tagged first). For signed PDFs, preserveSignatures: true repairs while keeping the signatures intact (approval signatures only; certification signatures are refused).
Inputs: inputPath (required), labels, outputPath, returnBase64, +2 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 20 tools
Each tool has a clearly distinct purpose: creation tools differ by input format, page manipulation tools (split, extract, delete, reorder) are well-separated, and annotation tools (bookmarks, annotations, watermark, page numbers) are unambiguous. Even tightly related operations like fill_form vs flatten_form are clearly delineated.
All 20 tool names follow a consistent verb_noun snake_case pattern (e.g., create_text_pdf, rotate_pages, set_metadata, ensure_pdfa). No camelCase or unpredictable verb forms appear, making the API highly predictable.
At 20 tools, this exceeds the typical 3-15 sweet spot and edges into 'borderline heavy', but each tool fills a distinct need within the broad PDF manipulation domain. The count is justified by the wide feature set (creation, page ops, forms, accessibility, archiving), so it earns its place despite being slightly over the ideal range.
The tool surface covers creation, page manipulation, metadata, bookmarks, annotations, forms, tagging, PDF/A, and attachments, which is comprehensive for a PDF writer. Minor gaps exist (e.g., image insertion, digital signing, or text extraction), but these are generally outside the stated purpose and do not cause dead ends in typical workflows.
Maintenance
Related MCP Connectors
Compliant PDFs (PDF/A-2A + PDF/UA-1) from markdown or a compact DSL - fast, no headless browser.
Extract text and metadata from PDFs; convert Markdown or text to PDF
Cloud PDF generation from HTML, CSS and XSL-FO, with PDF/A and PDF/UA support.
Render Markdown and LLM output into accessible PDF/UA-1 PDFs. No headless Chromium.
Related MCP Servers
- AlicenseAqualityDmaintenanceEnables PDF generation from HTML, text, and Markdown content with customizable formatting options. Provides secure cross-platform PDF creation tools that automatically save to user directories like Downloads, Documents, or Desktop.4MIT
- FlicenseNot gradedqualityDmaintenanceProvides tools for converting Markdown content and files into professional PDF documents with full support for Mermaid diagrams and LaTeX rendering. It allows for high-quality output customization, including paper size, table of contents, and syntax highlighting styles.-
- AlicenseNot gradedqualityDmaintenanceEnables PDF file manipulation including merging, splitting, extracting pages, extracting text, excluding pages, and reordering pages.1MIT
- AlicenseAqualityDmaintenanceEnables PDF processing and analysis including text extraction, metadata retrieval, search, page manipulation, splitting/merging, conversion to images, and form handling.10MIT