Snowflake MCP Server

snowflake-mcp-server : This server acts as bridge between your AI application and snowflake warehouse. This sever consist of various tools which will help you interact with the warehouse and perform operation seamlessly with minimal configuration.

MCP server List

** Snowflake MCP server ** Error_log MCP server -- for both the above server need to write the use case and advantages of using it.

What can be achived by using this tool

• Reduce the time between switching between your code IDE and Snowflake iterface
• Chat based database intereact
• This mcp server consist of in total 5 tools ** execute_query: This tool executes query on snowflake ** inspect_schema: This tool will inspect the table and return the description of the table. ** Analyse the performance: This tool can be used to analyse the performace of the query ** Check_data_quality: This tool will check the data quality in the table ** execute_batch: this toll will handel the multiple query run which is not handled by execute_query

*** This mcp server can handle both password based authentication and SSO authentication.

1. What is the MCP Server?

Model Context Protocol is an open standard that defines how models exchange, manage, and reason over shared context.

2. How MCP Servers Help Data Engineers

• MCP servers enable data engineers to connect AI models directly with enterprise data systems, streamlining access, automation, and insights.
• By providing standardized, secure connectivity between data warehouses and AI tools, MCP servers simplify integration and improve productivity.
• Data engineers can leverage MCP to automate data workflows, enhance context for AI applications, and reduce manual integration overhead.
• With MCP, data engineers spend less time managing connections and more time focusing on data quality, modeling, and innovation.

MCP architecture

It consist of three main part

1. MCP Host
2. MCP Client
3. MCP Server

3. Install as a Claude Code plugin

This repo is itself a Claude Code plugin: .claude-plugin/plugin.json and .mcp.json bundle both MCP servers (snowflake and snowflake-error-log). No cloning, and no manual pip install — .mcp.json runs both servers via uv run --with-requirements, which installs dependencies into an ephemeral env automatically on first launch. The only prerequisite is having uv on PATH.

On any machine with Claude Code and uv installed:

claude plugin marketplace add Data-Engineering-MCP/snowflake-mcp-server
claude plugin install snowflake-mcp-server@snowflake-mcp-server-marketplace

Claude Code will prompt for your Snowflake account, user, database, and warehouse, plus either a password or an authenticator (default externalbrowser for SSO). Sensitive values are stored in your OS keychain, not in plaintext settings. Re-run /plugin configure snowflake-mcp-server@snowflake-mcp-server-marketplace any time to change them.

To try local changes before pushing, use claude --plugin-dir /path/to/snowflake-mcp-server instead of installing from the marketplace.

4. MCP Config Section (manual / Cursor)

The server config below is for CURSOR IDE, or any MCP client that isn't Claude Code. Make sure you create env and install the required dependencies Below is an example of the configuration structure.

{
  "mcpServers": {
    "snowflake": {
      "command": "~/venv/bin/python3.13", 
      "args": ["path to main.py"]
    }
  }
}

1. Clone the repo
2. create a venv

python -m venv venv

1. Activate the venv

#for mac
source venv/bin/activate
#for windows
venv/Scripts/activate

1. Install all the dependencies present in requirements.txt

pip install -r requirements.txt

1. Setup the mcp server

• go to cursor settings -> mcp tool --> add server

1. Add the following JSON

{
  "mcpServers": {
    "error_log": {
      "command": "yourPath/snowflake_mcp_server/venv/bin/python3.13",
      "args": ["yourPath/snowflake-mcp-server/error_log_mcp/main.py"]
    }
  }
}

1. Change the path to point to the main.py and python in venv/bin or venv/scripts