概要
What is local_pgsql?
local_pgsql is a database analytics and query automation toolkit that provides structured data access for analysis, reporting, and AI-augmented workflows. It is designed for developers and data analysts who need to explore PostgreSQL databases, run parameterized SQL queries, and perform statistical analysis through a modular tool interface.
How to use local_pgsql?
Initialize the MCP with a database connection string, register tools and prompts using the provided API, then call any of the built-in tools or prompts by name. For example, register a DatabaseManager for a PostgreSQL URI and enable tools such as list_tables, run_query, or get_summary_statistics.
Key features of local_pgsql
- List all tables and get their schemas
- Preview table contents with data sampling
- Execute parameterized SQL queries
- Compute numerical summaries and correlation matrices
- Perform group-by aggregations with multiple functions
- Run temporal aggregation and anomaly detection
Use cases of local_pgsql
- Data exploration to rapidly understand dataset structure and content
- Automated reporting with scheduled statistical summaries
- Anomaly monitoring for real-time data quality checks
- AI-augmented analysis via structured data access for LLMs
FAQ from local_pgsql
What database systems does local_pgsql support?
The toolkit includes a DatabaseManager interface for backend-agnostic operations, and the integration example shows a PostgreSQL connection string.
What tools are available for statistical analysis?
Tools include get_summary_statistics (mean, standard deviation), analyze_correlations (correlation matrices), and group_by_analysis (multi-function aggregations).
Does local_pgsql include built-in prompts?
Yes, it provides a basic SQL guide (basic_sql_guide) and analysis task templates (data_analysis_tasks).
How are tools registered and invoked?
Tools are dynamically registered via the @mcp.tool decorator and called using mcp.call_tool with the tool name and parameters.
What type of anomaly detection is supported?
It offers Z-score and IQR based anomaly detection through the detect_anomalies tool.
