Data.gov MCP — wraps Data.gov CKAN API (catalog.data.gov/api/3)
Three read-only tools with clear verb-noun naming (search_datasets, get_dataset, list_organizations). All tools have descriptions (avg 150 chars, within baseline 194). Input schemas present with type definitions and required fields. However, parameter descriptions lack actionable constraints (no enums, ranges, or format specs). Output schemas are not documented, LLMs cannot predict response structure for chaining. Error handling is absent; API failures will surface raw HTTP errors. No pagination guidance despite search_datasets returning 20 results. Missing tool annotations (readOnlyHint). Composition is sound: each tool has one responsibility, and search_datasets returns organization names that match list_organizations output.
Get full metadata for a specific Data.gov dataset by its ID or name. Returns title, description, resources (download links), organization, tags, and update frequency.
List all organizations publishing datasets on Data.gov. Returns organization names and dataset counts. Useful for discovering data publishers.
Search U.S. government open datasets on Data.gov by keyword. Returns titles, descriptions, organizations, formats, and download URLs. Example: search_datasets("climate change", organization: "noaa-gov").
Output schemas not documented. LLMs cannot predict response structure (field names, types, nesting) for downstream tool calls or data extraction. searchDatasets returns {total_count, datasets[{id, name, title, description, organization, formats, modified}]} but this is invisible to the agent.
Parameter descriptions lack actionable constraints. 'organization' accepts 'noaa-gov', 'epa-gov', 'nasa-gov' but no enum or discovery mechanism. 'tags' is free-form string with no valid values listed. LLMs will hallucinate invalid values.
No error handling guidance. ckanGet() throws raw 'Data.gov API error: {status}' or '{message}'. LLMs receive no recovery hints (retry? lookup? ask user?). Example: 'Dataset not found' should suggest 'Try search_datasets() first.'
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2026-07-28+ | v2 |
No pagination or result-limit documentation. search_datasets hardcodes rows=20 but description does not state this limit. If CKAN returns 500 results, agent sees only 20 with no next_cursor or offset guidance.
Tool annotations missing. No readOnlyHint on any tool, despite all being read-only. Agents cannot distinguish safe tools from destructive ones without explicit hints.