PostgreSQL MCP server provides 7 read-only tools with basic schemas and minimal descriptions. All tools have input schemas with proper JSON Schema structure (type, additionalProperties, required arrays). However, descriptions are extremely terse (10-34 chars, well below the 50-200 char LLM-optimized baseline). No parameter descriptions exist, LLMs cannot infer what 'schema' or 'sql' parameters control. Output schemas are undocumented. Error handling is absent from visible code. Tools follow verb_noun naming (postgres_get_*, postgres_list_*) but descriptions lack WHEN/WHY context. No tool annotations (readOnlyHint, etc.) despite all being read-only. Composition is sound (each tool does one thing), but parameter naming could be more explicit (e.g., 'schema_name' vs 'schema').

• Expand all tool descriptions to 50-200 characters. Example: 'postgres_query' → 'Execute a read-only SELECT query against the connected PostgreSQL database. Returns structured rows. Use this to fetch data after discovering tables with postgres_list_tables and postgres_describe_table.'
• Add descriptions to all parameters. Example: 'schema' → 'The PostgreSQL schema name (e.g., public, information_schema). Use postgres_list_schemas to discover available schemas.'
• Document output schemas for each tool. Example: postgres_list_tables should return {tables: [{name: string, type: 'table'|'view', columns: number}], total: number}.
• Add tool annotations: set readOnlyHint=true on all 7 tools to signal they are safe to retry and do not modify state.
• For postgres_query, add parameter constraints: document that only SELECT queries are allowed, reject INSERT/UPDATE/DELETE/DROP at runtime with a clear error message, and return a structured result set with column names and types.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry