Von Snowflake verwalteter MCP-Server¶

Übersicht¶

Modell-Kontextprotokoll (MCP) ist ein Open-Source-Standard, der AI-Agenten erlaubt, sicher mit Geschäftsanwendungen und externen Datensystemen wie Datenbanken und Inhaltsrepositorys zu interagieren. Mit MCP können Unternehmen Integrationsherausforderungen reduzieren und schnell Ergebnisse aus Modellen liefern. Seit seiner Einführung ist MCP zu einer grundlegenden Unterstützung für agentenbasierte Anwendungen geworden und bietet einen konsistenten und sicheren Mechanismus für den Aufruf von Tools und den Abruf von Daten.

The Snowflake-managed MCP server lets AI agents securely retrieve data from Snowflake accounts without needing to deploy separate infrastructure. You can configure the MCP server to serve Cortex Analyst, Cortex Search, and Cortex Agents as tools, along with custom tools and SQL executions on the standards-based interface. MCP clients discover and invoke these tools, and retrieve data required for the application. With managed MCP servers on Snowflake, you can build scalable enterprise-grade applications while maintaining access and privacy controls. The MCP server on Snowflake provides:

• Standardisierte Integration: Einheitliche Schnittstelle für die Erkennung und den Aufruf von Tools, in Übereinstimmung mit den sich schnell entwickelnden Standards.

• Umfassende Authentifizierung: In Snowflake integrierter OAuth-Service, der OAuth-basierte Authentifizierung für MCP-Integrationen ermöglichen soll.

• Robuste Governance: Rollenbasierte Zugriffssteuerung (RBAC) für MCP-Server und -Tools zur Verwaltung der Erkennung und des Aufrufs von Tools.

Weitere Informationen zum MCP-Lebenszyklus finden Sie unter Lebenszyklus. Ein Beispiel für eine MCP-Implementierung finden Sie unter Erste Schritte mit dem verwalteten Snowflake MCP-Server Quickstart.

MCP server security recommendations¶

Using multiple MCP servers without verifying tools and descriptions could lead to vulnerabilities such as tool poisoning or tool shadowing. Snowflake recommends verifying third-party MCP servers before using them. This includes any MCP server from another Snowflake user or account. Verify all tools offered by third-party MCP servers.

We recommend using OAuth as the authentication method. Using hardcoded tokens can lead to token leakage.

When using a Programmatic Access Token (PAT), set it to use the least-privileged role allowed to work with MCP. This will help prevent leaking a secret with access to a highly-privileged role.

Configure proper permissions for the MCP server and tools following the least-privilege principle. Access to the MCP Server does not give access to the tools. Permission needs to be granted for each tool.

Erstellen eines MCP-Serverobjekts¶

Erstellen Sie ein Objekt, indem Sie die Tools und andere Metadaten angeben. MCP-Clients, die sich nach der erforderlichen Authentifizierung mit dem Server verbinden, können diese Tools erkennen und aufrufen.

1. Navigieren Sie zu der gewünschten Datenbank und dem gewünschten Schema, um den MCP-Server darin zu erstellen.

2. So erstellen Sie den MCP-Server:

   CREATE [ OR REPLACE ] MCP SERVER [ IF NOT EXISTS ] <server_name>
     FROM SPECIFICATION $$
       tools:
         - name: "product-search"
           type: "CORTEX_SEARCH_SERVICE_QUERY"
           identifier: "database1.schema1.Cortex_Search_Service1"
           description: "cortex search service for all products"
           title: "Product Search"
   
         - name: "revenue-semantic-view"
           type: "CORTEX_ANALYST_MESSAGE"
           identifier: "database1.schema1.Semantic_View_1"
           description: "Semantic view for all revenue tables"
           title: "Semantic view for revenue"
     $$
   

   Copy

Snowflake unterstützt derzeit die folgenden Tooltypen:

• CORTEX_SEARCH_SERVICE_QUERY: Cortex Search Service-Tool

• CORTEX_ANALYST_MESSAGE: Cortex Analyst-Tool

• SYSTEM_EXECUTE_SQL: SQL execution

• CORTEX_AGENT_RUN: Cortex Agent tool

• GENERIC: tool for UDFs and stored procedures

The following examples show how to configure different tool types:

Analyst toolSearch toolSQL execution toolAgent toolCustom tool

Using the Analyst tool, your client can generate text from SQL. Use the following code to specify the tool configuration.

Bemerkung

The Snowflake-managed MCP server only supports using semantic views with Cortex Analyst. It does not support semantic models.

tools:
  - name: "revenue-semantic-view"
    type: "CORTEX_ANALYST_MESSAGE"
    identifier: "database1.schema1.Semantic_View_1"
    description: "Semantic view for all revenue tables"
    title: "Semantic view for revenue"

Copy

Use the following examples to create and configure custom tools using UDFs and stored procedures:

-- create a simple udf
CREATE OR REPLACE FUNCTION MULTIPLY_BY_TEN(x FLOAT)
RETURNS FLOAT
LANGUAGE PYTHON
RUNTIME_VERSION = '3.8'
HANDLER = 'multiply_by_ten'
AS
$$
def multiply_by_ten(x: float) -> float:
  return x * 10
$$;

SHOW FUNCTIONS LIKE 'MULTIPLY_BY_TEN';

-- test return json/variant
CREATE OR REPLACE FUNCTION CALCULATE_PRODUCT_AND_SUM(x FLOAT, y FLOAT)
RETURNS VARIANT
LANGUAGE PYTHON
RUNTIME_VERSION = '3.8'
HANDLER = 'calculate_values'
AS
$$
import json

def calculate_values(x: float, y: float) -> dict:
  """
  Calculates the product and sum of two numbers and returns them in a dictionary.
  The dictionary is converted to a VARIANT (JSON) in the SQL return.
  """
  product = x * y
  sum_val = x + y

  return {
      "product": product,
      "sum": sum_val
  }
$$;

-- test return list/array
CREATE OR REPLACE FUNCTION GET_NUMBERS_IN_RANGE(x FLOAT, y FLOAT)
RETURNS ARRAY -- Use ARRAY to explicitly state a list is being returned
LANGUAGE PYTHON
RUNTIME_VERSION = '3.8'
HANDLER = 'get_numbers'
AS
$$
def get_numbers(x: float, y: float) -> list:
  """
  Returns a list of integers between x (exclusive) and y (inclusive).
  Assumes x < y.
  """
  # Ensure x and y are treated as integers for range generation
  start = int(x) + 1
  end = int(y) + 1 # range() is exclusive on the stop value

  # Use a list comprehension to generate the numbers
  # The Python list will be converted to a Snowflake ARRAY.
  return list(range(start, end))
$$;

-- create a simple stored procedure
CREATE OR REPLACE PROCEDURE MULTIPLY_BY_TEN_SP(x FLOAT)
RETURNS FLOAT
LANGUAGE PYTHON
RUNTIME_VERSION = '3.8'
PACKAGES = ('snowflake-snowpark-python')
HANDLER = 'multiply_by_ten'
AS
$$
# The handler logic is identical to the UDF for a scalar return
def multiply_by_ten(x: float) -> float:
      return x * 10
$$;

-- test return json/variant
CREATE OR REPLACE PROCEDURE CALCULATE_VALUES_SP(x FLOAT, y FLOAT)
RETURNS VARIANT
LANGUAGE PYTHON
RUNTIME_VERSION = '3.8'
PACKAGES = ('snowflake-snowpark-python')
HANDLER = 'calculate_values'
AS
$$
# The handler logic is identical to the UDF for a VARIANT return
def calculate_values(x: float, y: float) -> dict:
      """
      Calculates the product and sum of two numbers and returns them in a dictionary.
      The dictionary is converted to a VARIANT (JSON) in the SQL return.
      """
      product = x * y
      sum_val = x + y

      return {
          "product": product,
          "sum": sum_val
      }
$$;

-- test return list/array
CREATE OR REPLACE PROCEDURE GET_NUMBERS_SP(x FLOAT, y FLOAT)
RETURNS ARRAY
LANGUAGE PYTHON
RUNTIME_VERSION = '3.8'
PACKAGES = ('snowflake-snowpark-python')
HANDLER = 'get_numbers'
AS
$$
def get_numbers(x: float, y: float) -> list:
      """
      Returns a list of integers between x (exclusive) and y (inclusive).
      The Python list will be converted to a Snowflake ARRAY.
      """
      # Ensure x and y are treated as integers for range generation
      start = int(x) + 1
      end = int(y) + 1 # range() is exclusive on the stop value

      # Use a list comprehension to generate the numbers
      return list(range(start, end))
$$;

CREATE MCP SERVER my_mcp_server
  FROM SPECIFICATION $$
    tools:
      - title: "Custom Tool 1"
        identifier: "EXAMPLE_DATABASE.AGENTS.MULTIPLY_BY_TEN"
        name: "multiply_by_ten"
        type: "GENERIC"
        description: "Multiplied input value by ten and returns the result."
        config:
          type: "function"
          warehouse: "COMPUTE_SERVICE_WAREHOUSE"
          input_schema:
            type: "object"
            properties:
              x:
                description: "A number to be multiplied by ten"
                type: "number"
      - title: "Custom Tool 2"
        identifier: "EXAMPLE_DATABASE.AGENTS.CALCULATE_PRODUCT_AND_SUM"
        name: "calculate_product_and_sum"
        type: "GENERIC"
        description: "Calculates the product and sum of two numbers and returns them in a JSON object."
        config:
          type: "function"
          warehouse: "COMPUTE_SERVICE_WAREHOUSE"
          input_schema:
            type: "object"
            properties:
              x:
                description: "First number"
                type: "number"
              y:
                description: "Second number"
                type: "number"
      - title: "Custom Tool 3"
        identifier: "EXAMPLE_DATABASE.AGENTS.GET_NUMBERS_IN_RANGE"
        name: "get_numbers_in_range"
        type: "GENERIC"
        description: "Returns a list of integers between two numbers."
        config:
          type: "function"
          warehouse: "COMPUTE_SERVICE_WAREHOUSE"
          input_schema:
            type: "object"
            properties:
              x:
                description: "Start number (exclusive)"
                type: "number"
              y:
                description: "End number (inclusive)"
                type: "number"
      - title: "Custom Tool 4"
        identifier: "EXAMPLE_DATABASE.AGENTS.MULTIPLY_BY_TEN_SP"
        name: "multiply_by_ten_sp"
        type: "GENERIC"
        description: "Multiplied input value by ten and returns the result."
        config:
          type: "procedure"
          warehouse: "COMPUTE_SERVICE_WAREHOUSE"
          input_schema:
            type: "object"
            properties:
              x:
                description: "A number to be multiplied by ten"
                type: "number"
      - title: "Custom Tool 5"
        identifier: "EXAMPLE_DATABASE.AGENTS.CALCULATE_PRODUCT_AND_SUM_SP"
        name: "calculate_product_and_sum_sp"
        type: "GENERIC"
        description: "Calculates the product and sum of two numbers and returns them in a JSON object."
        config:
          type: "procedure"
          warehouse: "COMPUTE_SERVICE_WAREHOUSE"
          input_schema:
            type: "object"
            properties:
              x:
                description: "First number"
                type: "number"
              y:
                description: "Second number"
                type: "number"
      - title: "Custom Tool 6"
        identifier: "EXAMPLE_DATABASE.AGENTS.GET_NUMBERS_IN_RANGE_SP"
        name: "get_numbers_in_range_sp"
        type: "GENERIC"
        description: "Returns a list of integers between two numbers."
        config:
          type: "procedure"
          warehouse: "COMPUTE_SERVICE_WAREHOUSE"
          input_schema:
            type: "object"
            properties:
              x:
                description: "Start number (exclusive)"
                type: "number"
              y:
                description: "End number (inclusive)"
                type: "number"
  $$;

1. Zum Anzeigen von MCP-Servern verwenden Sie die folgenden Befehle:

   SHOW MCP SERVERS IN DATABASE <database_name>;
   SHOW MCP SERVERS IN SCHEMA <schema_name>;
   SHOW MCP SERVERS IN ACCOUNT;
   

   Copy

   The following shows the output of the command:

   |               created_on               |       name        | database_name | schema_name |    owner     |           comment            |
   ------------------------------------------+-------------------+---------------+-------------+--------------+------------------------------
   | Fri, 23 Jun 1967 07:00:00.123000 +0000 | TEST_MCP_SERVER   | TEST_DATABASE | TEST_SCHEMA | ACCOUNTADMIN | [NULL]                       |
   | Fri, 23 Jun 1967 07:00:00.123000 +0000 | TEST_MCP_SERVER_2 | TEST_DATABASE | TEST_SCHEMA | ACCOUNTADMIN | Test MCP server with comment |
   

2. Um einen MCP-Server zu beschreiben, verwenden Sie den folgenden Befehl:

   DESCRIBE MCP SERVER <server_name>;
   

   Copy

   The following shows the output of the command:

   |      name       | database_name | schema_name |    owner     | comment |     server_spec        |               created_on               |
   ------------------------------------------------------------------------------------------------------+-------------------------------------
   | TEST_MCP_SERVER | TEST_DATABASE | TEST_SCHEMA | ACCOUNTADMIN | [NULL]  | {"version":1,"tools":[{"name":"product-search","identifier":"db.schema.search_service","type":"CORTEX_SEARCH_SERVICE_QUERY"}]} | Fri, 23 Jun 1967 07:00:00.123000 +0000 |
   

3. Um einen MCP-Server zu entfernen, verwenden Sie den folgenden Befehl:

   DROP MCP SERVER <server_name>;
   

   Copy

Zugriffssteuerung¶

Sie können die folgenden Berechtigungen verwenden, um den Zugriff auf das MCP-Objekt und die zugrunde liegenden Tools zu verwalten

Einrichten der OAuth-Authentifizierung¶

Konfigurieren Sie die Authentifizierung auf dem MCP-Client. Der von Snowflake verwaltete MCP-Server unterstützt OAuth 2.0 in Übereinstimmung mit der Empfehlung zur `Autorisierung<https://modelcontextprotocol.io/specification/2025-06-18/basic/authorization>`__ im MCP-Protokoll. Der von Snowflake verwaltete MCP-Server unterstützt keine dynamische Clientregistrierung.

1. Erstellen Sie zuerst die Sicherheitsintegration. Weitere Informationen zu diesem Befehl finden Sie unter CREATE SECURITY INTEGRATION (Snowflake OAuth).

   CREATE [ OR REPLACE ] SECURITY INTEGRATION [IF NOT EXISTS] <integration_name>
     TYPE = OAUTH
     OAUTH_CLIENT = GENERIC
     ENABLED = TRUE
     OAUTH_CLIENT_TYPE = 'CONFIDENTIAL'
     OAUTH_REDIRECT_URI = '<redirect_URI>'
   

   Copy

2. Rufen Sie dann die Systemfunktion auf, um Ihre Client-ID und die Schlüssel für die Clientkonfiguration abzurufen. Der Integrationsname unterscheidet zwischen Groß- und Kleinschreibung und muss in Großbuchstaben geschrieben werden.

   SELECT SYSTEM$SHOW_OAUTH_CLIENT_SECRETS('<integration_name>');
   

   Copy

Interagieren mit MCP-Server mit einem kundenspezifischen MCP-Client¶

Falls Sie einen kundenspezifischen MCP-Client erstellen, werden in den folgenden Abschnitten die Snowflake-Endpunkte beschrieben, mit denen Ihr Client interagiert.

POST /api/v2/databases/{database}/schemas/{schema}/mcp-servers/{name}
    {
        "jsonrpc": "2.0",
        "id": 1,
        "method": "initialize",
        "params": { "protocolVersion": "2025-06-18" }
    }

{
    "jsonrpc": "2.0", // passthrough from req
    "id": 1, // passthrough from req
    "result": {
        "proto_version": "2025-06-18",
        "capabilities": {
            "tools": {
                "listChanged": false
            }
        },
        "server_info": {
            "name": "<snowflake-mcp-name>",
            "title": "Snowflake Server: <snowflake-mcp-name>",
            "version": "1.0.0"
        }
    }
}

POST /api/v2/databases/{database}/schemas/{schema}/mcp-servers/{name}
    {
        "jsonrpc": "2.0",
        "id": 1,
        "method": "tools/list",
        "params": {}
    }

{
    "jsonrpc": "2.0", // passthrough from req
    "id": 1, // passthrough from req
    "result": {
        "tools": [ // search
            {
                "name":"product-search",
                "description":"Test search tool",
                "inputSchema": {
                    "type": "object",
                    "description": "A search query and additional parameters for search.",
                    "properties": {
                        "query": {
                            "description": "Unstructured text query.  Exactly one of 'query' or 'multi_index_query' must be specified.",
                            "type": "string"
                        },
                        "columns": {
                            "description": "List of columns to return.",
                            "type": "array",
                            "items": {
                                "type": "string"
                            }
                        },
                        "filter": {
                            "description": "Filter query.",
                            "type": "object"
                        },
                        "limit": {
                            "description": "Max number of results to return.",
                            "type": "integer",
                            "default": 10
                        }
                    }
                },
                "outputSchema": {
                    "type": "object",
                    "description": "Search results.",
                    "properties": {
                        "results": {
                            "description": "List of result rows.",
                            "type": "array",
                            "items": {
                                "type": "object",
                                "additionalProperties": true,
                                "description": "Map of column names to values (as bytes)."
                            }
                        },
                    "request_id": {
                        "description": "ID of the request.",
                        "type": "string"
                    }
                },
                "required": ["results", "request_id"]
            }
          },
          { // analyst
              "name":"revenue-semantic-view",
              "description":"Test tool",
              "inputSchema": {
                  "type": "object",
                  "description": "A message and additional parameters for Cortex Analyst.",
                  "properties": {
                      "message": {
                          "description": "The user’s question.",
                          "type": "string"
                      }
                  }
              }
          }
        ]
    }
}

POST /api/v2/databases/<database>/schemas/<schema>/mcp-servers/<name>
    {
        "jsonrpc": "2.0",
        "id": 1,
        "method": "tools/call",
        "params": {
            "name": "test-analyst",
            "arguments": {
                "message ": "text"
            }
        }
    }

{
    "jsonrpc": "2.0",
    "id": 1,
    "result": {
        "content": [
            {
                "type": "text",
                "text": "string"
            }
        ]
    }
}

POST /api/v2/databases/{database}/schemas/{schema}/mcp-servers/{name}
    {
        "jsonrpc": "2.0",
        "id": 1,
        "method": "tools/call",
        "params": {
            "name": "product-search",
            "arguments": {
                "query": "Hotels in NYC",
                "columns": array of strings,
                "filter": json,
                "limit": int
            }
        }
  }

{
    "jsonrpc": "2.0",
    "id": 1,
    "result": {
        "results": {}
    }
}

Einschränkungen¶

Der Snowflake-verwaltete MCP-Server unterstützt die folgenden Konstrukte im MCP-Protokoll nicht: Ressourcen, Prompts, Roots, Benachrichtigungen, Versionsverhandlungen, Lebenszyklusphasen und Sampling.

Es werden nur Nicht-Streaming-Antworten unterstützt.