This server exposes 4 tools (list_metrics, get_dimensions, get_entities, query_metrics) for dbt Semantic Layer integration via HTTP. The code is visible and tool definitions can be extracted from streamlit_v1_single_sl.py. However, critical gaps emerge: (1) Tool descriptions are present but minimal (10-30 chars, well below the 50-200 char LLM-optimized baseline). (2) Parameter descriptions are sparse or absent for most inputs, e.g., 'where' parameter has no explanation of SQL syntax expectations, 'order_by' lacks format guidance. (3) No output schemas are documented, the code extracts results via extract_text() and json.loads(), but the expected response structure is never formalized for the LLM. (4) Error handling is present but generic ('Could not read content', 'Non-JSON data'), no recovery guidance (e.g., 'Try calling list_metrics first'). (5) All tools are READ_ONLY, which is good, but this constraint is not formally annotated in tool definitions (no toolAnnotations/readOnlyHint per the spec). The tool names follow verb_noun convention correctly, but parameter design lacks constraints and validation guidance.

• Document output schema for each tool. Example: list_metrics returns { metrics: [{ name: string, description: string, type: string }], total: number }. Include this in the tool description so the LLM knows what to extract.
• Expand descriptions to 50-200 characters each. Example: 'list_metrics' → 'List all available metrics from the dbt Semantic Layer. Call this first to discover metric names before querying. Returns a paginated list of metric objects with name, description, and type.'
• Add parameter constraints. For query_metrics: 'where' → 'Optional SQL WHERE clause. Example: "revenue > 1000 AND date >= 2024-01-01". Standard SQL syntax.' 'order_by' → 'Array of objects. Each has { field: string (metric or dimension name), direction: "asc"|"desc" }.' 'limit' → 'Maximum rows to return. Must be 1 - 10000. Default: 100.'
• Add recovery guidance to error messages. Instead of 'Could not read content from list_metrics', return 'Error parsing metrics catalog. Ensure the dbt Semantic Layer is accessible and your authentication token is valid. Try calling list_metrics again, or check your MCP_URL and DBT_TOKEN.'
• Formally annotate tools with readOnlyHint: true in the tool definition. This tells the agent these are safe to call multiple times without side effects.
• For get_dimensions and get_entities, clarify the dependency: 'Must be called after list_metrics() to identify valid metric names. Returns a list of dimension/entity names available for grouping or filtering these metrics.'
• Add parameter validation examples to descriptions. For 'metrics' array, state: 'Example: ["revenue", "active_users"]. Names are case-sensitive and must exist in the Semantic Layer.'