WildBerries Data MCP Server

MCP (Model Context Protocol) server for WildBerries data that allows an AI to directly query the database using natural language. The server provides a secure interface for AI to query the database while ensuring query safety through the QueryGuard package.

Features

• Provides an MCP-compliant API for AI-driven database queries
• Connects to both ClickHouse and PostgreSQL databases
• Ensures query security through QueryGuard validation
• Returns results in JSON format for easy consumption by AI

Prerequisites

• Go 1.18+
• ClickHouse database
• PostgreSQL database
• Redis (for caching)

Environment Variables

Configure the application using the following environment variables:

# Server configuration
WB_DATA_MCP_IP=0.0.0.0
WB_DATA_MCP_PORT=8081

# Clickhouse configuration
CH_USERNAME=default
CLICKHOUSE_PASSWORD=your_password
CLICKHOUSE_DATABASE=your_database
CH_HOST=localhost
CH_PORT=9000

# PostgreSQL configuration
PSQL_USERNAME=postgres
PG_HOST=localhost
PG_PORT=5432

# Redis configuration
REDIS_HOST=localhost
REDIS_PORT=6379
REDIS_PASSWORD=
REDIS_DB=0

# Telegram logging
TELEGRAM_CHAT_ID=your_chat_id
TELEGRAM_BOT_TOKEN=your_bot_token

Usage

1. Clone the repository:

git clone https://github.com/marketconnect/wb_data_mcp.git

2. Set environment variables (see above)

3. Build the application:

go build -o wb_data_mcp

4. Run the server:

./wb_data_mcp

Query Example

The server provides a query_data tool that AI can use to query the database. Here's an example request:

{
  "table_name": "stocks",
  "fields": ["product_id", "warehouse_id", "quantity", "basic_price"],
  "filters": [
    {"field": "product_id", "operator": "=", "value": "123456"}
  ],
  "limit": 10
}

Allowed Tables

The server supports querying the following tables:

• stocks: Product stock information
• orders: Order history
• orders30d: 30-day order aggregation
• subjects: Subject categories
• Various other PostgreSQL tables defined in the system

Security

QueryGuard ensures that only safe, read-only queries can be executed, protecting your database from potentially harmful operations. The system enforces:

• Only SELECT statements are allowed
• Only pre-approved tables and columns can be queried
• No modifications (INSERT, UPDATE, DELETE) are permitted
• No dangerous SQL constructs (subqueries, CTEs, etc.)

License

This project is licensed under the MIT License