MCP server for storing and retrieving conversation memory
Two tools with minimal quality. Tool 1 ('Add or Update Note') has a vague name combining two actions (add AND update), lacks clarity on when each applies, and has no output schema documentation. Tool 2 ('search_notes') uses a cryptic parameter name 'ss' instead of 'search_term' or 'query'. Both descriptions are under 100 chars and lack context on when to use them. Parameter descriptions are minimal (10-20 chars). No error handling guidance, errors return raw exception messages. No pagination support for search results. No input validation or constraints documented.
Add a new note to the wiki or update an existing note
Search for notes in the wiki
Tool name 'Add or Update Note' combines two distinct actions (add AND update). Should split into separate tools: 'create_note' and 'update_note' so the agent can compose them intentionally.
Parameter 'ss' is cryptic and non-standard. Rename to 'search_term' or 'query' to match LLM expectations and improve discoverability.
No output schema documented. LLMs cannot plan downstream calls or extract structured data. Document what search_notes returns (e.g., list of note objects with id, title, snippet).
Error handling returns raw exception messages ('Error adding note: ' + error.message). Provide actionable recovery guidance: 'Note name already exists. Use update_note() instead.' or 'Wiki service unavailable. Retry in 30 seconds.'
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | 2026-07-28+ | v2 |
search_notes returns raw JSON dump of response.data with no pagination, result count, or limit. Large result sets will blow context window. Add limit parameter (default 20) and return structured results with total_count.