MCP server for querying Sofia municipality open urban data from the CKAN portal
Sofia Data MCP demonstrates solid fundamentals: all 11 tools have clear verb-noun naming (search_, get_, list_, preview_, summarize_, find_, facet_), descriptions between 34 - 194 chars, and explicit Zod schemas with typed parameters. All parameters have descriptions. However, output schemas are not documented in the source, only inferred from handler return types. Error handling is minimal; no recovery guidance or actionable error messages visible. Tool descriptions lack dependency hints and prerequisites. No tool annotations (readOnlyHint, etc.) despite all tools being read-only. Pagination is present in search_datasets but not consistently applied across list tools.
Return catalog-wide counts grouped by group, organization, format, and license, computed by CKAN across all matching datasets (not just a page of results)
Find related datasets by shared groups, tags, publisher, and formats
Get a dataset by ID or name
Get a group by ID or name
Get an organization by ID or name
List resources for a dataset
List dataset groups
Output schemas not documented. Handler return types are inferred but not explicitly declared in tool definitions. LLMs cannot plan downstream calls without knowing what fields to expect.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All 11 tools are read-only but this is not declared in the tool definition, forcing LLMs to infer safety from descriptions alone.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 69 | <=2025-11-25 | v2 |
List publishing organizations
Preview a dataset resource when the format is text-friendly
Search datasets in the Sofia urban data portal
Summarize a dataset including formats, external hosts, and inferred data meaning when available
Minimal error handling. No recovery guidance, actionable error messages, or error classification (retryable vs user-fixable vs fatal). Handlers do not validate input or provide helpful failure context.
Inconsistent pagination. search_datasets supports rows/start pagination, but list_groups, list_organizations, and list_dataset_resources do not. Large result sets will blow context windows.
Descriptions lack dependency hints and prerequisites. E.g., preview_resource does not explain that datasetId and resourceId must come from prior calls to get_dataset and list_dataset_resources.