MCP server for retrieving context from a Qdrant vector database
Two tools with basic schemas and descriptions, but significant gaps in parameter documentation and output structure. qdrant-find has a vague description that doesn't explain WHEN to use it vs alternatives. qdrant-store lacks clarity on metadata format. Neither tool documents return structure or pagination. Parameters lack type constraints (e.g., query_filter is 'object' with no schema). No error handling guidance. Naming is acceptable (verb_noun pattern) but descriptions fall short of the 50-200 char LLM-optimized baseline.
Look up memories in Qdrant. Use this tool when you need to: - Find memories by their content - Access memories for further analysis - Get some personal information about the user
Keep the memory for later use, when you are asked to remember something.
query_filter parameter is typed as 'object' with no schema definition. LLMs cannot infer valid filter structure, leading to malformed queries.
qdrant-find description is vague ('Look up memories...') and does not explain when to use it, what it returns, or how results are structured. Lacks WHAT/WHEN/RETURN clarity.
No output schema documented for either tool. qdrant-find returns list[str] but structure of entries is undocumented. qdrant-store returns string but format is unclear.
metadata parameter in qdrant-store accepts 'any json' but no constraints or examples guide LLM on valid structure. Risk of malformed metadata.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 51 | <=2025-11-25 | v2 |
No error handling guidance. If collection_name does not exist or query fails, LLM has no recovery path. Missing recovery-guide pattern.