Snowflake MCP Server

Instructions

Implementation Reference

• src/mcp_snowflake_server/server.py:183-221 (handler)

  The main handler function that parses the fully qualified table name, executes a query against INFORMATION_SCHEMA.COLUMNS to retrieve column metadata (name, default, nullable, data_type, comment), formats the output as YAML text and JSON embedded resource.

  async def handle_describe_table(arguments, db, *_):
      if not arguments or "table_name" not in arguments:
          raise ValueError("Missing table_name argument")
  
      table_spec = arguments["table_name"]
      split_identifier = table_spec.split(".")
  
      # Parse the fully qualified table name
      if len(split_identifier) < 3:
          raise ValueError("Table name must be fully qualified as 'database.schema.table'")
  
      database_name = split_identifier[0].upper()
      schema_name = split_identifier[1].upper()
      table_name = split_identifier[2].upper()
  
      query = f"""
          SELECT column_name, column_default, is_nullable, data_type, comment 
          FROM {database_name}.information_schema.columns 
          WHERE table_schema = '{schema_name}' AND table_name = '{table_name}'
      """
      data, data_id = await db.execute_query(query)
  
      output = {
          "type": "data",
          "data_id": data_id,
          "database": database_name,
          "schema": schema_name,
          "table": table_name,
          "data": data,
      }
      yaml_output = data_to_yaml(output)
      json_output = json.dumps(output)
      return [
          types.TextContent(type="text", text=yaml_output),
          types.EmbeddedResource(
              type="resource",
              resource=types.TextResourceContents(uri=f"data://{data_id}", text=json_output, mimeType="application/json"),
          ),
      ]

• src/mcp_snowflake_server/server.py:464-478 (registration)

  Registers the describe_table tool in the all_tools list, specifying name, description, input schema (requiring fully qualified table_name), and linking to the handle_describe_table handler. This tool object is then filtered and exposed via list_tools and call_tool.

  Tool(
      name="describe_table",
      description="Get the schema information for a specific table",
      input_schema={
          "type": "object",
          "properties": {
              "table_name": {
                  "type": "string",
                  "description": "Fully qualified table name in the format 'database.schema.table'",
              },
          },
          "required": ["table_name"],
      },
      handler=handle_describe_table,
  ),

• src/mcp_snowflake_server/server.py:467-476 (schema)

  Defines the input schema for the describe_table tool: an object with a required 'table_name' string property expecting 'database.schema.table' format.

  input_schema={
      "type": "object",
      "properties": {
          "table_name": {
              "type": "string",
              "description": "Fully qualified table name in the format 'database.schema.table'",
          },
      },
      "required": ["table_name"],
  },