A document indexing and search server with MCP (Model Context Protocol) integration, supporting multiple document sources (local, GitHub, GitLab, HTTP) with semantic and keyword search capabilities
DocSanctum exposes 4 read-only tools via FastMCP with HTTP transport. Tool names follow verb_noun convention (list_, read_, search_, semantic_search_). However, critical gaps undermine quality: no input schemas are visible in the provided source code, parameter descriptions are absent, and output schemas are undocumented. Descriptions are present but generic (10-60 chars), lacking LLM-optimized guidance on when to use each tool vs. alternatives. The semantic_search_documents tool lacks clarity on query format and embedding model behavior. No error handling guidance, no pagination support visible, and no indication of result limits. Tools appear functional but lack the structured metadata production systems require.
List all indexed documents across sources
Read the content of a specific document
Search documents by keyword
Search documents by semantic meaning using a natural-language query
No input schemas visible in source code. All 4 tools lack documented parameter types, constraints, and descriptions.
Tool descriptions are generic and under 60 characters. 'List all indexed documents across sources' lacks guidance on when to call this vs. search_documents, what structure is returned, or pagination behavior.
semantic_search_documents description does not explain query format, embedding model, or similarity threshold. LLMs cannot infer whether to pass natural language, keywords, or structured queries.
No output schemas documented. LLMs cannot plan downstream calls or extract required fields (e.g., document_id for read_document). Baseline: 100% of A+ tools have documented return types.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 46 | <=2025-11-25 | v2 |
No pagination, result limits, or error handling guidance visible. list_documents and search_documents may return unbounded results, risking context window exhaustion.