A Model Context Protocol server exposing a notes knowledge base as tools + resources.
Three tools with complete schemas and descriptions. Naming follows verb_noun convention (search_notes, add_note, get_note). Descriptions are present but brief (10-60 chars), below the 50-200 char LLM-optimized baseline. Parameter descriptions exist but lack detail on constraints, formats, and error recovery. Output schemas are implicit in code (text responses) rather than formally documented. Error handling is minimal, get_note returns isError flag but lacks actionable recovery guidance. No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Search result limiting is good (1-20 range), but no pagination support for large result sets.
Add a new note to the knowledge base.
Fetch a single note by id.
Search the knowledge base for notes matching a query.
Descriptions are too brief (10-60 chars vs 50-200 char baseline). 'Search the knowledge base for notes matching a query' lacks WHEN to use, prerequisites, or what happens on no results.
Parameter descriptions lack constraint details. 'limit' says 'optional limit between 1 and 20' but should explain default (5), when to increase, and impact on response time. 'query' lacks guidance on format (keywords, phrases, regex?).
No tool annotations. add_note should have destructiveHint=true (modifies state), search_notes and get_note should have readOnlyHint=true. Agents cannot infer safety/idempotency without these hints.
Error handling lacks recovery guidance. get_note returns isError=true with 'No note with id {id}' but does not suggest search_notes or list available notes. Agents cannot self-correct.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | 2026-07-28+ | v2 |
Output schemas are implicit (text responses in code) rather than formally documented. LLMs cannot plan downstream calls or extract structured data. search_notes returns text with '#id [score] title' format but no schema definition.