MCP server: generate professional PDFs from JSON in Claude, Cursor, Windsurf, or any AI agent. Serverless, no Chromium. Invoices, reports, resumes.
Six tools with well-structured schemas and detailed descriptions. Naming is clear and action-oriented (generate_*, list_*, validate_*). Descriptions are comprehensive (100 - 300 chars), exceeding the 10 - 1024 baseline. Input schemas are fully specified with types, enums, and nested objects. However, output schemas are not documented, LLMs cannot predict return structure. Error handling is present but lacks recovery guidance. No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Parameter descriptions are thorough but some nested object properties lack inline documentation.
Convert a Markdown string to a PDF. Supports headings, bold/italic, strikethrough, inline code, links, ordered/unordered lists (nested up to 3 levels), GFM tables (with column alignment), GFM task lists (☑/☐), blockquotes, fenced code blocks (rendered as plain indented text — not styled monospace), and horizontal rules. For richer layouts or styled code blocks, use generate_pdf instead. Returns a base64-encoded PDF.
Generate a professional invoice PDF. Accepts structured invoice data (from/to parties, line items, optional GST/tax). Returns base64-encoded PDF. Supports INR/USD/EUR/GBP currencies. Currency symbols (₹ $ € £) are guaranteed not to break away from adjacent numbers across line wraps. gst_rate on items creates a tax column — use it for any tax system (GST, VAT, sales tax). Per-item gst_rate is summed as IGST in the totals; CGST/SGST inter-vs-intra-state routing is not currently supported.
Generate a PDF from a custom PdfDocument JSON descriptor. Use this for any layout not covered by generate_invoice or generate_report — e.g. resumes, contracts, certificates, presentations, or any multi-element composition. Returns base64-encoded PDF bytes. Call list_element_types first to see available elements and options.
Generate a professional multi-section report PDF. Features: auto-generated TOC, per-section headings with PDF bookmarks, page-numbered footer, running header, optional tables and callout boxes per section. Returns base64-encoded PDF.
Output schemas not documented. LLMs cannot predict return structure (base64 PDF bytes, validation errors, element list format). Agents cannot plan downstream operations or extract fields reliably.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All tools are read-only (no side effects), but this is not declared. Agents cannot distinguish safe tools from destructive ones without explicit hints.
Error handling lacks recovery guidance. PretextPdfError exceptions are caught but responses do not suggest next steps (e.g., 'Call list_elements() to see available element types' when an invalid element is used).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 63 | 2026-07-28+ | v2 |
List all available PdfDocument element types and their schemas. Use this to discover what content elements you can use in generate_pdf.
Validate a PdfDocument JSON structure without rendering. Returns validation errors if the document is malformed or contains unsupported element types.
generate_pdf description references 'Call list_element_types first' but the tool is named 'list_elements'. Naming mismatch creates confusion for LLMs.
Nested object properties (e.g., margins.top, header.fontSize, encryption.permissions) lack inline descriptions. LLMs must infer meaning from property names alone.