A MCP server for DuckDB that enables LLMs to query and analyze data from various sources
This server has critical gaps in definition quality. While all 5 tools are present with basic descriptions and input schemas, the descriptions are minimal (mostly 50-100 chars), parameter descriptions are sparse or missing entirely, output schemas are undocumented, and there is no error handling guidance. The tools follow a verb_noun pattern (create_session, query, analyze_schema, analyze_data, suggest_visualizations), which is good, but descriptions lack actionable context for LLM selection. Parameters like 'file_path' and 'table_name' in analyze_data and suggest_visualizations have basic descriptions but do not explain mutual exclusivity, valid formats, or when to use each. No output schema is documented, forcing LLMs to infer response structure. Error paths return bare TextContent without recovery hints. The server has potential for data analysis workflows but lacks the specificity and guardrails expected of production tools.
Perform statistical analysis on data in a table or file, including row counts, numeric statistics, date ranges, and categorical value distributions
Analyze the schema of a file or table, returning column definitions and types
Create a new session for managing user state and query history
Execute a SQL query against the DuckDB database
Generate visualization suggestions for a table or file based on column types and data characteristics
No output schemas documented. LLMs cannot infer response structure, making it impossible to chain tools or extract specific fields from results.
Parameter descriptions are minimal or absent. 'file_path' and 'table_name' in analyze_data and suggest_visualizations lack detail on format, mutual exclusivity, valid examples (e.g., S3 paths, file extensions), or when to use each.
Tool descriptions are too short (35 - 60 chars) to provide LLM-actionable context. Descriptions should explain WHAT, WHEN to use, and prerequisites (10 - 1024 chars baseline; 50 - 200 chars is ideal for LLM clarity). Current descriptions are 40% of recommended minimum.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 53 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 46 | - | v1 |
No error handling or recovery guidance. Handlers return bare 'Error: ...' TextContent without actionable next steps. LLMs cannot distinguish retryable errors from user-fixable issues.
Parameter type definitions incomplete. 'query' input schema shows only name/description; 'table_name' and 'file_path' lack explicit type in schema (inferred as string from description only).
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). While all tools appear read-only, explicit annotation would clarify safety guarantees for agents.