MCP server for personal knowledge base management with advanced RAG capabilities, vector search, document listing, and knowledge base metrics
PersonalKB_Server has three tools with basic descriptions and minimal parameter documentation. Tool names follow verb_noun convention (search_*, list_*), which is good. However, all three tools lack input parameter descriptions, the 'query' parameter in search tools has no guidance on format, length, or expected content. Output schemas are undocumented; responses are free-text strings rather than structured objects. Error handling is present but generic ('An error occurred...') without recovery guidance. No tool annotations (readOnlyHint, etc.) despite all being read-only. Descriptions are adequate (100-200 chars) but lack WHEN-to-use clarity that distinguishes the two search tools for LLM selection.
List the names of all documents stored in the personal knowledge base. Use this when the user asks what documents, files, specs, or logs are available for search.
Search the knowledge base using advanced two-stage retrieval (Vector Search + Cross-Encoder Re-ranking). Use this when the user asks specific technical questions, queries about specifications, metrics, configurations, or architectures that may be in their documents.
Search the knowledge base using raw vector similarity search. This does NOT perform re-ranking or relevance gate filtering. Use this when you want a fast, simple keyword match.
No input parameter descriptions. 'query' parameter in search tools lacks format guidance, length constraints, or examples of when to use raw vs. advanced search.
Output schemas undocumented. All tools return free-text strings; no structured schema defined. LLMs cannot plan downstream operations or extract typed fields.
Generic error messages without recovery guidance. Errors like 'An error occurred during raw search: {e}' do not tell LLM what to do next or whether to retry.
No tool annotations. All three tools are read-only but lack readOnlyHint annotation. Agents cannot infer safety/idempotency from tool metadata.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 52 | <=2025-11-25 | v2 |
Ambiguous tool selection. Both search tools have similar names and descriptions; LLM reasoning required to distinguish 'raw' vs. 'advanced'. Add explicit WHEN-to-use guidance.