Pure-Rust MCP server for PDF intelligence: read, search, document analysis with page-level proof, structured text, tables, OCR, visual evidence, and citations
Six tools with basic descriptions and partial schemas. All tools have descriptions (10-100 chars, mostly 50-90 chars), but parameter descriptions are minimal or absent. Input schemas visible but lack depth: 'sources' and 'query' params have descriptions, but no type constraints (enums, min/max, patterns). No output schemas documented. Error handling not visible in tool definitions. Tools follow verb_noun naming (read_pdf, search_pdf, pdf_compare) but some names are redundant (search_pdf vs pdf_text_search both search; pdf_evidence overlaps with search_pdf). No evidence of idempotency hints, permission gates, or recovery guidance in descriptions.
Compare two PDF documents for structural and content differences
Extract visual evidence and citations from PDF documents with bounding boxes and page references
Compute cryptographic hash of PDF documents for integrity verification
Search PDF text content with advanced indexing and matching
Read and extract structured content from PDF documents with text, tables, OCR, and visual analysis
Search PDF documents for text patterns with page-level citations and bounding box evidence
No output schemas documented. LLMs cannot plan downstream calls or extract structured data without knowing what fields to expect.
Redundant tool names: search_pdf and pdf_text_search both search text; pdf_evidence overlaps with search_pdf functionality. LLMs will waste reasoning cycles deciding between them.
Parameter descriptions lack constraints. 'sources' accepts 'file paths or URLs' but no format validation, length limits, or error guidance. 'pages' accepts 'optional page numbers' but no min/max bounds (what if user passes page 999999?).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2026-07-28+ | v2 |
No error handling guidance in tool descriptions. If a PDF is corrupted, unreadable, or a URL is unreachable, LLMs have no recovery path documented.
Tool descriptions do not explain when to use each tool instead of similar ones. Why call pdf_evidence instead of search_pdf? What does 'visual evidence' mean vs 'text search'?