A comprehensive Model Context Protocol (MCP) server engineered to bridge the gap between technical database schemas and AI-driven intelligence. This server enables AI agents to extract, enrich, and visualize complex database architectures across various platforms including PostgreSQL, Snowflake, SQL Server, BigQuery, and others.

Architecture Overview

The system operates as a comprehensive six-stage pipeline, transforming raw technical specifications into business-ready intelligence and structured canonical models.

Architecture Overview

Core Capabilities

Deep Schema Extraction

Automatically scan and retrieve metadata for tables, columns, indexes, constraints, and sequences. Supports a wide array of database engines with secure credential management.

AI-Driven Metadata Enrichment

Leverages advanced Agent Intelligence to transform cryptic technical names into meaningful business descriptions. Automatically infers missing primary and foreign key relationships to reconstruct complex data models.

Interactive Intelligence Reports

Generates high-fidelity HTML reports powered by D3.js. These reports include interactive relationship graphs, searchable schema tables, and detailed documentation for every database entity.

Canonical Data Modeling (CDM)

Generate industry-standard Canonical Data Models directly from business requirements using AI. Supports domain grouping, audit column standardisation, and relationship inference.

Semantic Mapping & Lineage

Automated mapping of source metadata to CDM entities. Tracks lineage, transformations, and provides confidence scores for every field-level mapping.

Native VS Code Integration

Built for the Model Context Protocol, allowing seamless interaction within VS Code Agent Mode.

Infrastructure Requirements (Optional)

While the "version": "0.1.8", will attempt to create these automatically if they don't exist, you may manually initialize the following directory structure in your project root to support the end-to-end metadata pipeline:

mkdir raw_json enriched_json html_output

raw_json/: Destination for technical schema dumps.
enriched_json/: Repository for AI-augmented metadata.
html_output/: Hosting folder for interactive visual reports.

Configuration (.env)

Secure your credentials by creating a .env file with the following parameters:


# Connectivity
DATABASE="orc"
USERNAME="your_username"
PASSWORD="your_password"

Installation

VS Code Marketplace

Navigate to the Extensions panel in VS Code.
Search for "DB Metadata Extractor MCP".
Select Install.
Verify Python 3.10+ is available on your system path.

Operational Workflow

Once the server is connected to your AI agent, you can initiate tasks using simplified natural language commands:

Step 1: Extract

"Extract metadata from my Oracle database and save the output to raw_json/."

Step 2: Enrich

"Run the metadata-enrichment skill on the latest file in raw_json/ and save to enriched_json/."

Step 3: Visualize Schema

"Generate an interactive HTML report from the enriched JSON file and save to html_output/."

Step 4: CDM Generation

"Use the cdm-generation skill to design a Canonical Data Model for a Waste Management company with Customer and Generator domains, saving to cdm_output/."

Step 5: Semantic Mapping

"Use the mapping skill to map the enriched source JSON to the new CDM, saving to mapping_output/."

Step 6: Visualize Mapping

"Generate a mapping report from the CDM and mapping JSON files and save to html_output/ using the generate_mapping_report tool."

Complete Pipeline (Using .env Credentials)

"Perform the full end-to-end pipeline: Extract database metadata, enrich it, and design a Waste Management CDM. Then map the source metadata to the CDM and generate both the schema and mapping reports in html_output/."