This server has critical definition gaps. Tool descriptions are present but generic (50-100 chars, below the 194-char baseline). Input schemas are visible but lack depth: parameters have types and basic descriptions, but no enums, ranges, or format constraints. The archive_document tool exposes user_token as a parameter, a critical security violation. No output schemas are documented. Error handling is absent from descriptions. Tool naming is acceptable (verb_noun pattern), but composition issues exist: archive_document combines file upload, OCR wait, and AI metadata analysis, three concerns that should be separate. No pagination support visible for search_documents despite returning lists.

• Remove user_token from all tool parameters. Inject it server-side from environment or config. Update tool signatures to accept only user_id or infer from request context.
• Split archive_document into three tools: (1) upload_document(file_bytes, file_name) → doc_id; (2) wait_for_ocr(doc_id, timeout_seconds=30) → ocr_status; (3) analyze_and_set_metadata(doc_id) → {title, tags, correspondent, storage_path}. This lets agents handle OCR delays and failures independently.
• Document output schemas for all tools. E.g., search_documents returns {total_count: int, results: [{doc_id: int, title: str, tags: [str], correspondent: str, download_url: str}], next_offset: int}.
• Add pagination to search_documents: limit (1-100, default 20) and offset (default 0). Return total_count and next_offset so agents can iterate.
• Expand descriptions to 150-250 chars. Include WHEN to call, what it returns, and error recovery hints. E.g., get_document: 'Downloads a document by Paperless ID and returns it as a PDF file. Use after search_documents to retrieve full content. Returns binary PDF data. Fails if doc_id is invalid or user lacks read permission.'

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry