MCP server for searching and retrieving datasets from the Borealis Dataverse repository, a Canadian research data repository.
Single-tool server with a well-structured input schema and clear descriptions. The search_datasets tool has comprehensive parameter documentation with enums, defaults, and explanatory text. However, no documented output schema is visible in the source code, and the server lacks error handling guidance and recovery patterns. The tool description is strong (170+ chars) with actionable guidance about DOI/author inclusion, but the implementation does not appear to validate that guidance or provide structured error feedback to guide LLM self-correction.
Search for datasets in Borealis Dataverse by keywords, subjects, or other criteria. Returns a list of matching datasets with basic information. When presenting results to the user, ALWAYS include the full DOI URL (e.g., https://doi.org/10.34990/FK2/XXXXX) and the authors for each dataset in your summary. Do not omit author information.
No documented output schema. The tool description and input schema are well-defined, but the expected response structure (fields, types, pagination info) is not visible in the source. LLMs cannot predict what data they will receive or plan downstream tool calls without seeing the output format.
No error handling guidance or recovery patterns. If the API returns a 404 (dataverse not found), authentication failure, rate limit, or malformed query, there is no visible code to translate those failures into actionable LLM guidance (e.g., 'Try a different dataverse name', 'Use uppercase AND/OR/NOT for boolean operators').
Boolean operators constraint documented only in description text. The 'query' parameter description explains 'use uppercase AND, OR, NOT', but this is not enforced in the schema (no regex pattern, no enum validation). If the user passes lowercase 'and', the tool may silently treat it as a literal keyword instead of guiding the LLM to correct the case.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 59 | 2026-07-28+ | v2 |
Dataverse and Country/Province filters rely on description examples ('University of Toronto', 'Canada') but no enum or validation visible in schema. LLMs may hallucinate invalid institution names or misspelled province codes, causing silent failures or empty result sets.