NitinTalluri/mcp_server

LLM with Claude - Snowflake Architecture MCP Server

A Python project that provides an MCP (Model Context Protocol) server for accessing Snowflake architecture data tables. This server enables LLM models to query live architecture data as a source of truth for application structure analysis, dependencies, and relationships.

Features

• MCP server for Snowflake architecture data access
• Direct querying of architecture tables through LLM tools
• Schema inspection and data search capabilities
• Integration with Claude models via Continue IDE extension
• Support for multiple environments (dev, stage, prod)

Architecture Tables Available

The server provides access to these key architecture tables:

Core Entity Tables

• BASE_NAMES - Source table naming information
• VIEW_NAMES - View table naming information
• BASE_TABLE - Source table definitions with columns
• VIEW_TABLE - View table definitions with columns

Lineage & Dependencies

• VIEW_TO_SOURCE_COLUMN_LINEAGE - Raw column lineage data
• VIEW_TO_SOURCE_COLUMN_LINEAGE_V - Enhanced column lineage view

Application Integration

• ACTION_TO_ENDPOINTS_TABLES_MAPPING - Raw endpoint to table mappings
• ACTION_TO_ENDPOINTS_TABLES_MAPPING_V - Enhanced endpoint mappings

Business Intelligence

• THOUGHTSPOT_TABLE - ThoughtSpot table usage data
• THOUGHTSPOT_TABLE_V - Enhanced ThoughtSpot integration view

Stored Procedures

• SP_TABLE_COLUMN_MAPPING - Stored procedure metadata
• SP_TABLE_COLUMN_MAPPING_V - Enhanced SP mappings
• SP_REFERENCE_TO_FLOWS - Flow reference data
• SP_MAPPING_WITH_ENDPOINTS_AND_FLOWS_V - Comprehensive SP mapping

Setup Guide

Prerequisites

• Python 3.10 or higher
• Access to Snowflake environment
• Continue IDE extension (for LLM integration)

Installation

1. Clone the repository

   git clone <repository-url>
   

2. Create virtual environment

   python -m venv .venv
   

3. Activate virtual environment

   Windows:

   .venv\Scripts\activate
   

   Linux/Mac:

   source .venv/bin/activate
   

4. Install dependencies

   pip install -r requirements_mcp.txt
   

   Or using uv (if available):

   uv sync
   

Configuration

1. Snowflake Connection Setup

   • Ensure you have access to Snowflake environments
   • The server uses existing connection strategy via common.sec module
   • Default environment is set to 'prod' but can be modified in mcp_snowflake_server.py

2. Continue IDE Configuration

   • The project includes config_snowflake.yaml for Continue IDE replcae this with your config.yaml
   • Update the cwd path in the config to match your project location

Running the MCP Server

1. Direct execution

   python mcp_snowflake_server.py
   

2. Via Continue IDE

   • The server will automatically start when Continue IDE loads the configuration
   • Check the MCP server configuration in config_snowflake.yaml

Usage

Available MCP Tools

1. query_architecture_table

   • Query specific tables with optional filtering and limits
   • Parameters: table_name (required), limit (optional), where_clause (optional)

2. get_table_schema

   • Get column information and schema details for tables
   • Parameters: table_name (required)

3. search_architecture_data

   • Search across tables for specific patterns or relationships
   • Parameters: search_term (required), tables (optional)

Troubleshooting

Common Issues

1. Connection Errors

   • Verify Snowflake credentials and access
   • Check network connectivity
   • Ensure the common.sec module is properly configured

2. MCP Server Not Starting

   • Verify Python path in config_snowflake.yaml replcae this in config.yaml
   • Check virtual environment activation
   • Review server logs for specific errors

3. Missing Dependencies

   • Run pip install -r requirements_mcp.txt
   • Ensure all required packages are installed