Snowflake MCP Server

Overview

A Model Context Protocol (MCP) server implementation that provides database interaction with Snowflake. This server enables running SQL queries with tools and intereacting with a memo of data insights presented as a resource.

Components

Resources

The server exposes a single dynamic resource:

• memo://insights: A continuously updated data insights memo that aggregates discovered insights during analysis
  • Auto-updates as new insights are discovered via the append-insight tool

Tools

The server offers six core tools:

Query Tools

• read_query

  • Execute SELECT queries to read data from the database
  • Input:
    • query (string): The SELECT SQL query to execute
  • Returns: Query results as array of objects

• write_query (with --allow-write flag)

  • Execute INSERT, UPDATE, or DELETE queries
  • Input:
    • query (string): The SQL modification query
  • Returns: { affected_rows: number }

• create_table (with --allow-write flag)

  • Create new tables in the database
  • Input:
    • query (string): CREATE TABLE SQL statement
  • Returns: Confirmation of table creation

Schema Tools

• list_databases

  • Get a list of all databases in the Snowflake instance.
  • No input required
  • Returns: Array of database names.

• list_schemas

  • Get a list of all schemas in a specific database.
  • Input:
    • database (string): Name of the database.
  • Returns: Array of schema names.

• list_tables

  • Get a list of all tables in a specific database and schema.
  • Input:
    • database (string): Name of the database.
    • schema (string): Name of the schema.
  • Returns: Array of table metadata.

• describe-table

  • View column information for a specific table
  • Input:
    • table_name (string): Fully qualified name of table to describe (e.g., database.schema.table)
  • Returns: Array of column definitions with names and types

Analysis Tools

• append_insight
  • Add new data insights to the memo resource
  • Input:
    • insight (string): data insight discovered from analysis
  • Returns: Confirmation of insight addition
  • Triggers update of memo://insights resource

Local Usage with Claude Desktop or Cursor

Below is a quick guide to get started, and a more general and detailed guide can be found here.

1. Install Claude AI Desktop App or Cursor

2. Install uv by:

curl -LsSf https://astral.sh/uv/install.sh | sh

3. Create a .env file using the following template under this dir

SNOWFLAKE_USER="[email protected]"
SNOWFLAKE_ACCOUNT="XXX"
SNOWFLAKE_ROLE="XXX"     # This determines the access scope of the MCP
SNOWFLAKE_DATABASE="XXX" # This doesn't affect the MCP's access scope
SNOWFLAKE_SCHEMA="XXX"   # This doesn't affect the MCP's access scope
SNOWFLAKE_WAREHOUSE="XXX"
SNOWFLAKE_AUTHENTICATOR="externalbrowser"  # Use Okta for authentication

4. [Optional] Modify the exclude_patterns in runtime_config.json to filter out the resources you want to exclude.

5. Test locally using

uv --directory /absolute/path/to/mcp_snowflake_server run mcp_snowflake_server

6. To use Claude Desktop: add the server to your claude_desktop_config.json (Claude -> Settings -> Developer -> Edit Config), then restart Claude Desktop.

"mcpServers": {
  "snowflake_local": {
      "command": "/absolute/path/to/uv", # obtained by using `which uv`
      "args": [
          "--directory",
          "/absolute/path/to/mcp_snowflake_server",
          "run",
          "mcp_snowflake_server",
          # Optionally: "--allow_write" (but not recommended)
          # Optionally: "--log_dir", "/absolute/path/to/logs"
          # Optionally: "--log_level", "DEBUG"/"INFO"/"WARNING"/"ERROR"/"CRITICAL"
          # Optionally: "--exclude_tools", "{tool name}", ["{other tool name}"]
      ]
  }
}

7. To use Cursor: add the following to your Curser Settings -> MCP -> Add new MCP server -> Command

/absolute/path/to/uv --directory /absolute/path/to/mcp_snowflake_server run mcp_snowflake_server